Yukiko Tsuboi Draft of Qual Paper

Total Page:16

File Type:pdf, Size:1020Kb

Yukiko Tsuboi Draft of Qual Paper

Yukiko Tsuboi Draft of Qual Paper

Is there "Raising" in Japanese?

1. Introduction: Raising

English is known to have an operation called raising, which dislocates John in the embedded subject position in (0a) to the matrix subject position as in (0b), and Mary in the embedded subject position in (0a) to, the matrix object position as in (0b).

(0) a. It seems that John is a genius. b. John seems to be a genius. (0) a. John believes that Mary is a genius. b. John believes Mary to be a genius.

(0b) is said to correspond to (0b) in Japanese, in which John with the nominative case-marker in (0a) is marked with the accusative case-marker.

(0) a. Mary-ga [John-ga bakada-to] omotteiru (koto) Mary-NOM [John-NOM stupid:be-that] think (that)

'(that) Mary thinks that John is stupid' b. Mary-ga John-o bakada-to omotteiru (koto) Mary-NOM John-ACC stupid:be-that think (that)

'(that) Mary thinks of John that he is stupid'

(0) can be schematized as in (0);

(0) a. ...NP1-NOM...[NP2- NOM...V2-to]...V1 (that)

b. ...NP1- NOM...NP2-ACC...V2-to...V1 (that)

In a sentence of the form in (0a), an NP in the embedded subject, that is, NP2 in (0), can optionally

1 Yukiko Tsuboi Draft of Qual Paper

be marked with the accusative marker in (0b). Let us call a sentence whose schematic structure corresponds to (0b) Raising-to-Object sentence (RTO), and NP2-o in (0b), NPAcc.

It is note that there are two restrictions on the type of the matrix verb; (i) V1 must be a verb which can take an embedded clause headed by the "complementizer", -to 'that', and (ii) V1 is a verb which does not take an accusative marked NP only under the same interpretation of V1 in (0a). The verb occurs as V1 in (0), omotteiru, does not take an accusative marked NP solely without CP complement under the interpretation of 'think' in (0). When it does, it is interpreted as 'to care about somebody, to have feelings for somebody' as in (0) and (0).

(0) a. John-ga Mary-o omotteiru (koto) John-NOM Mary-ACC think (that)

'John has feelings for Mary.' / 'John cares about Mary.' b. [Oya-ga kodomo-o omou kimoti]-wa umi-yori fukai. Parent-NOM child-ACC think feelings-TOP sea:from deep

Lit. 'The feelings that the parent cares about the child is deeper than the depth of the sea.'

(0) a. John-wa Mary-no kenkoo-o omotte, diet-o susumeta. John-TOP Mary-GEN health-ACC think:and, diet-ACC suggest

'Since he cares about her health, John suggested that Mary be on a diet.' b. [musume-no syoorai-o omotte] John-wa singaku-o susumeta. daughter-GEN future-ACC think:and John-TOP higher:education-ACC gave:uprecommended(?)

'John recommended his daughter that she pursue a higher education because he cares about her future.'

As illustrated in (0) and (0), omotteiru 'to think' cannot be interpreted as 'to think about somebody to be something' as in the case of (0) when it takes an accusative marked NP only. There are four types of analyses proposed for RTO constructions in Japanese: (a) Raising analysis (proposed in Kuno 1976, and further defended in Sakai 1998, Yoon 2004), (b) Major Object analysis (hinted in Saito 1983, explicitly proposed in Hong 1990 and Hoji 1991, and further defended in Takano 2003), (c) ECM analysis (proposed in J.E.Yoon 1989) and (d) Combination

2 Yukiko Tsuboi Draft of Qual Paper

analysis (Hiraiwa 2002, Bruening 2001). Under the Raising analysis, NPAcc is analyzed as being generated in the embedded clause and raising to the matrix clause. Major Object approach argues that NPAcc is base-generated in the matrix clause and never be a part of the embedded clause. The ECM analysis claims that NPAcc is base-generated within the embedded clause and stays inside of the embedded CP throughout the derivation. The combined analysis of the Raising and the ECM analysis proposes that there are two types of RTO in Japanese; in one type, NPAcc undergoes raising and in the other, it stays in the embedded clause throughout the derivation. In this paper, I will defend [the base-generation analysis of RTO<=Major Object analysis のこ とですよね? ] by showing invalidity of the arguments provided in support of the raising or the ECM analysis. In Section 2, I will summarize the Raising Analysis and the Major Object Analysis and evaluate arguments provided in support for each analysis. In section 3, the ECM analysis will be discussed in depth and two major arguments, based on final-particle attachment and "indeterminate agreement,", are evaluated. Following Takano 2002 and Harada 2002, I provide counterexamples to the generalization put forth for the "indeterminate agreement" phenomenon and claim that ["indeterminate agreement" phenomenon is a result of special constituency formation proposed by Kawazoe 2004 <==Not all instances of "indeterminate agreement" involve special constituency formation proposed by Kawazoe 2004, does it?] .

2. Raising Analysis and Major Object Analysis

Both the raising analysis and the Major Object analysis claim that NPAcc is an element of the matrix clause at some level of derivation. The only difference is whether or not NPAcc is considered as originating in the embedded clause. In this section, we will first summarize claims of each analysis and illustrate structures they propose. In Section 2.3, we discuss one type of arguments provided by the Raising analysis advocates which attempts to show that NPAcc behaves as if it is an element of the matrix clause and [show<=これは、どこにかかりますか? わかりにく かったです。we discuss ,,, and show ですか、それとも which attempt(s) ... and show ですか?] that they are not valid. In Section 2.4, we provided two other arguments of this type which convincingly show that NPAcc belongs to the matrix clause. [It is noted(?) that this type of argument could be a support for the Major Object analysis since it claims that NPAcc is in the matrix clause throughout the derivation.<=section 2 の目的はそもそも summarize claims of each analysis (=raising analysis and Major Object anaysis)なのですよね? そう思って読んできたら、 この"NPAcc belongs to the matrix clause. It is noted that...”ときたので、ちょっと面食らいまし

3 Yukiko Tsuboi Draft of Qual Paper

た。Section 2 全体が、Raising analysis のみに関することだとしたら、「Raising の evidence とし て出されているけれども」、という前振りに続いて It is noted that でも OK だとは思いますが 。] [The other type of arguments ---there is only one argument of this type<=? the other で、there is only one argument ですか?]---given by the Raising analysis is the PBC violation effects which are meant to show that NPAcc originates in the matrix clause. The validity of the argument crucially distinguishes the Raising analysis from the Major Object analysis since the PBC effects can only be accounted for by the former analysis which assume a movement of NPAcc from the embedded clause to the matrix clause. Based on the results of surveys, we discuss the PBC effects extensively in Section 2.5 and argue that the alleged PBC effects observed in RTO is not as robust as "real PBC effects" known in the literature.

2.1 Raising Analysis Kuno 1976 first claims that RTO construction in Japanese involves raising of the embedded subject which can be stated as in (0).

(0) NPAcc is generated in the embedded CP and gets raised to the matrix clause.

Under the raising analysis, two types of structures have been proposed; (i) NPAcc is generated as the subject of the embedded clause claimed in Kuno 1976, further defended in Sakai 1998, and (ii) NPAcc is generated as Major Subject proposed in Yoon 2004, 2005 and to appear. (0) is an illustration of the structure proposed by the former analysis. Under this analysis, the NPAcc is base-generated in a position where the embedded subject is generally assumed to be generated, and it gets raised to the position of the matrix object.

(0) ...NP1- NOM...NP2-ACC...V2-to...V1 (that) TP wo vP T wo

NP1-NOM v' ty

4 Yukiko Tsuboi Draft of Qual Paper

VP ty

NP2 i- ACC V' (=NPAcc) ty

CP V1 ty TP -to ty vP wo

ti. v' 5

...V2 Under the analysis pursued in Yoon 2004, 2005 and to appear, it is proposed that NPAcc is base-generated in the position of Major Subject of the embedded clause. Following the proposal by Heycock 1993, Doron & Heycock 1999, Yoon 1987, 2001, Yoon assumes that Major Subject occupies higher Specifier(s) of TP.

(0) ...NP1- NOM...NP2-ACC...V2-to...V1 (that)

TP wo vP T wo

NP1-NOM v' ty VP ty

[NP2-Acc]i V' (=NPAcc) ty

CP V1 ty TP -to

5 Yukiko Tsuboi Draft of Qual Paper

ty

[Major Subject ti.] T' ty vP T 5

...V2

The arguments provided in support of the raising analysis can be divided into two types; the first type is meant to show that NPAcc behaves as if it is an element of the matrix clause, and the second type is intended to show that NPAcc originates in the embedded clause. The detail of each argument will be discussed in [the following three sections<=2.2 は違うので、starting from 2.3 と 書いてあるといいかと思いました。].

2.2 Major Object Analysis Major Object Analysis, which is explicitly proposed in Hoji 1991 in the framework of the Generative Grammar, and further defended in Takano 2003, claims that NPAcc is base-generated in the matrix clause and never a part of the embedded clause throughout the derivation. (0) illustrates the proposed structure.

(0) ...NP1- NOM...NP2-ACC...V2-to...V1 (that) TP wo vP T wo

NP1-NOM v' ty VP ty

NP2 i- ACC V' (=NPAcc) ty

CP V1 ty TP -to

6 Yukiko Tsuboi Draft of Qual Paper

ty vP wo

proi / ec v' 5

...V2

Hoji 1991 further claims that NPAcc is an adjunct corresponding to of NP in English, the essence of which is followed by Takano 2003. Their analyses differ in optionally of existence of pro inside the embedded clause being co-indexed with NP-Acc. We will not discuss this issue in this paper1 since the existence or non-existence of a pro inside the embedded clause does not affect the discussion here.

2.3 Invalid Arguments for NPAcc being in Matrix Clause In this sub-section, we will evaluate three empirical observations listed in (0), all of which are provided to show that NPAcc behaves like an element of the matrix clause.

(0) a Adverb Placement b. Word Order Inversion c. Quantifier Scope

We will show that all of the observations above cannot convincingly show that NPAcc is in the matrix clause.

2.3.1Adverb Placement Kuno 1976 observes that the adverbs modifying matrix verbs cannot be preceded by the embedded subject as in (0a) while it can be preceded by NPAcc as in (0b).

1 The acceptable sentence in (i) seems to suggest that pro in the embedded clause is not obligatory. (i) Mary-ga John-o a-itu-wa sekaiiti-no baka da to omotteiru (koto) Mary-NOM John-ACC that:guy-TOP workd:#1-GEN stupid be-C is:thinking (that) 'Mary is thinking about John that that guy is the most stupid guy in the world.'

7 Yukiko Tsuboi Draft of Qual Paper

(0) Kuno 1976 25: (21d), (22d) a. *Yamada wa [Tanaka-ga, orokanimo, tensaida ]to omotte ita Yamada-TOP [Tanaka-NOM, foolishly, genius:be]C think past

Intended: Yamada foolishly thought that Tanaka is a genius.' b. Yamada wa Tanaka-o, orokanimo, tensaida to omotte ita Yamada-TOP Tanaka-ACC, foolishly, genius:be C think past

In (0a), Tanaka, which is the embedded subject, precedes the matrix adverb, orokanimo, and the sentence is reported to be unacceptable in Kuno 1976. (0b) is an acceptable RTO sentence, in which Tanaka is NPAcc preceding the matrix adverb. The generalization put forth can be stated as in (0).

(0) (i) The matrix adverb cannot be preceded by the embedded subject. (ii) NPAcc can be preceded by the matrix adverb.

The reported contrast in (0) can be taken as evidence that Tanaka-o in (0b) is in the matrix clause under the assumptions that (0) are both correct.

(0) a. Adverbs cannot move downward, or b. Subject NPs cannot be scrambled.

Let us assume that an adverb is adjoined to VP as is generally assumed, the structure of which is illustrated in (0). In Example (0a), orokanimo is adjoined to the matrix VP preceding the embedded clause. If the embedded subject in (0a) moves out of the embedded clause crossing over the matrix adverb, we will have (0b);

(0) a. Yamada wa [VPorokanimo [VP [S Tanaka-ga tensaida to] omotte ita]

b. Yamada wa Tanakai-ga orokanimo ti tensaida to omotte ita

Notice that (0b) and (0b) are phonetically identical. If the reported contrast in (0) is correct, Kuno 1976 will have to disallow scrambling of the subject NP, which is argued to be possible in Ko 2004. [Ko 2004 reported that the following Korean sentence is acceptable unless parsing difficulty

8 Yukiko Tsuboi Draft of Qual Paper

arises. (0) is a corresponding Japanese example which is also acceptable under the intended reading.

(0) Ko 2004: 2: (7)

John-ii [na-nun [ti Mary-lul cohahanta-ko] sayngkakhanta] John-NOM I-TOP Mary-ACC like-C think

'John thinks that I like Mary.'

(0) John-ga [watasi-wa [ti Mary-o sukida] to omotteiru. John-NOM I-TOP Mary-ACC like C think

The observation above shows that one of the assumptions in (0) is not correct, which cast doubt on validity of the generalization in (0i), repeated below.<=この部分、確か Saito 1985 の Chap. 4 だっ たと思うのですが、Subject は Movement の対象にはならないというところで、(0)のような例 が OK なのは Topic-NP の lowering が OK だからだという議論があったような気がします。 うろ覚えで、しかも今手元にないので確認できません、すみません。]

(0) (i) The matrix adverb cannot be preceded by the embedded subject.

Let us look at (0) again, the schematic structures of which is given in (0).

(0) a. NP1-NOM...NP2-NOM...Adv...{NP-be / Adjective}-to...V2 (predicted to be out by Kuno 1976)

b. NP1-NOM...NPAcc...Adv...{NP-be / Adjective}-to...V2

The sentences of the form in (0ab) are, in fact, not unacceptable to many of the native speakers. E. Mukai conducted an experiment (Fall 2004) to check the generalization in (0) and the result shows that there is not sharp contrast between sentences of the forms in (0a) and (0b) although (0a) is somewhat degraded to some speakers. It is noted(?) that the degraded status of (0a) does not support Kuno's claim because (0b<=a?) should never be generated under the grammar he proposes. If he were to maintain the

9 Yukiko Tsuboi Draft of Qual Paper

generalization in (0) with the observation that the subject can scramble, he will have to make an ad hoc stipulation, for instance, that(?) the embedded subject NP cannot scramble crossing a matrix adverb but the matrix subject NP can scramble crossing a topic phrase in the matrix clause. In summary, the generalization in (0i) is not valid, and it cannot be a support for the claim that NPAcc is in the matrix clause.

2.3.2Word Order Inversion Another observation provided by Kuno in support for NPAcc being in the matrix clause is that NPAcc behaves different from the embedded subject with respect to availability of scrambling. (0a) is a simplex sentence in which three NPs, a topic NP, the dative NP and the accusative NP are ordered canonically, and (0b) illustrates that the accusative NP can scramble over the topic.

(0) Kuno 1976: 26: (26) a. Yamada-wa Morita-ni Tanaka-o syookaisita. Yamada-TOP Morita-DAT Tanaka-ACC introduced

'Yamada introduced Tanaka to Morita.' b. Tanaka-o Yamada-wa Morita-ni syookaisita. Tanaka-ACC Yamada-TOP Morita-DAT introduced

In the case of (0), a sentence is embedded inside the object phrase and the embedded subject cannot be dislocated to sentence-initial position as shown in (0b).

(0) Kuno 1976: 26: (27) a. Yamada-wa [Tanaka-ga tensai dearu]koto-o siranakatta. Yamada-TOP Tanaka-NOM genius be fact-ACC know:not:past

'Yamada did not know that Tanaka was genius.' b. *Tanaka-ga, Yamada-wa tensaidearu koto-o siranakatta.

Based on the observation above, the following generalization is made;

(0) The element in the embedded clause cannot be dislocated to the matrix clause.

10 Yukiko Tsuboi Draft of Qual Paper

Given (0), Kuno observes the contrast between (0) and (0), both of which are complex sentences.

(0) Kuno 1976: 26: (28) a. Yamada-wa Tanaka-ga tensaida to omotteita. Yamada-TOP Tanaka-Nom genius be C thought

'Yamada thought that Tanaka was genius.' b. *Tanaka-ga Yamda-wa tensaida to omotteita.

In (0a), Tanaka is the embedded subject and it cannot be dislocated to sentence initial position as in (0b). In the case of RTO construction in (0), on the other hand, NPAcc, Tanaka, can be dislocated without affecting acceptability of the sentence as in (0b) contra to (0b).

(0) Kuno 1976: 26: (29) a. Yamada-wa Tanaka-o tensaida to omotteita. Yamada-TOP Tanaka-ACC genius be C thought

b. Tanaka-o, Yamada-wa tensaida to omotteita.

The acceptable status of (0b) is taken as evidence that Tanaka-o is not a constituent of the embedded clause but a constituent of the matrix clause. This observation cannot be correct because scrambling of subject NPs is allowed in grammar as we have seen in the previous section. One may wonder what the source of the unacceptable status of (0b) is. As noted in Ko 2004, availablity / unavaliability of scrambling of subject NPs is affected largely by parsing difficulty. If we supply enough contexts to parse sentences, (0b) can easily be accepted;

(0) Michiko de-wha nakute, Tanaka-ga, Yamda-wa tensaida to omotteita. Michiko be-TOP not:and, Tanaka-NOM Yamada-TOP genius be C think

'Yamada thinks that Tanaka, not Michiko, is genius."

In (0), Michiko dewha nakute 'not Michiko' is added to the sentence in (0b) by making Tanaka-ga "contrastive", and the sentence is fairly acceptable and the generalization in (0) is not valid. Thus,

11 Yukiko Tsuboi Draft of Qual Paper

the word-order observation cannot constitute evidence for the position of NPAcc.

2.3.4 Parallelism between the accusative object in the matrix clause and NPAcc The third argument for the location of NPAcc is provided based on scope interaction between NPAcc and the matrix subject. Kuno reports that NPAcc in RTO behaves in the same way as the accusative NP in a simplex clause as illustrated in (0) and (0).

(0) Kuno 1976: 28: (35) & 27: (32)

a. [NP1Dareka-ga] [[NP2 minnna-ga] sinda]koto-o siranakatta. NP1 >NP2, *NP2>NP1

someone-NOM everyone-NOM died fact-ACC know:not:past

'Someone did knot know everyone died.'

b. [NP1Dareka-ga] [NP2 minnna-o] aisiteiru. NP1 >NP2, NP2>NP1 someone-NOM everyone-ACC loves

'Someone loves everyone.'

In the complex sentence in (0), the embedded subject cannot take a wide scope with respect to the matrix subject while the object can take scope over the subject in the simplex sentence in (0b). The same contrast is said to exhibit in the complex sentences in (0), (b) exmaple of which is a RTO sentence.

(0) Kuno 1976: 29: (37) & (38)

a. [NP1Dareka-ga] [[NP2 minnna-ga] baka da] to omotteiru. NP1 >NP2, *NP2>NP1 someone-NOM everyone-NOM stupid be C think

'Someone thinkis that eveyone is stupid.'

b. [NP1Dareka-ga] [NP2 minnna-o] baka da-to omotteiru. NP1 >NP2, NP2>NP1 Someone-NOM everyone-ACC stupid be-C thinks

'Someone thinks of everyone to be stupid.'

In (0), the embedded subject cannot but NPAcc can take a wide scope with respect to the matrix subject.

Ueyama 1998 claims and Hayashisita 1999, 2004 extensively discuss that when a QP 1 takes a

12 Yukiko Tsuboi Draft of Qual Paper

wide scope with respect to a QP2, the former must c-command the latter. Hayasisita 1999, 2004 further argues that inverse scope yields not through Grammar but through "predication". In order to avoid scope interpretation based on "predication", a certain type of QPs must be used.

(0) Ueyama 1998: 124: (12) A-type QPs ex. NP-sae 'even NP' kanarinokazu-no NP 'most of the NPs' 10 izyoo-no NP' 'ten or more NPs' 55%-no NP '55% of the NPs' NP1 to NP2 (to) 'NP1 and NP2' NP1 ka NP2 (ka) 'either NP1 or NP2'

Observe (0), sentences with A-type QPs

(0) a. [NP1 sanninizyoo-no sensei-ga] [[NP2 futariizyoo-no gakusei-ga] ryuugakusita to more:than:three-GEN teacher-NOM more:than:two-GEN student-NOM study:abroad:past C

siranakatta (koto) know:neg:past (fact)

'More than three teachers did not know that more than two students went to study

abroad.' NP1 >NP2, *NP2>NP1

b. [NP1 sanninizyoo-no sensei-ga] [NP2 futariizyoo-no gakusei]-o ryuugakusita to more:than:three-GEN teacher-NOM more:than:two-GEN student-ACC study:abroad:past C

siranakatta (koto) know:neg:past (fact)

NP1 >NP2, *NP2>NP1

The scope ambiguity no longer observes between NP1 and NP2, and the same is true of (0).

13 Yukiko Tsuboi Draft of Qual Paper

(0) a. [NP1 sanninizyoo-no sensei-ga] [[NP2futariizyoo-no gakusei]-ga yuusyuuda]to omotteiru more:than:three-GEN teacher-NOM more:than:two-GEN student-NOM smart:be C think

'More than three teachers thinks that more than two students are smart.'

(koto) NP1 >NP2, *NP2>NP1 (fact)

b. [NP1sanninizyoo-no sensei-ga] [NP2 futariizyoo-no gakusei]-o yuusyuuda to omotteiru more:than:three-GEN teacher-NOM more:than:two-GEN student-ACC smart:be C think

(koto) NP1 >NP2, *NP2>NP1 (fact)

Based on the observation above, we can conclude that NPAcc as well as the embedded subject cannot take scope over the matrix subject, which merely suggests that NPAcc cannot be in the position which c-commands the matrix subject. Thus, the contrast does not show whether NPAcc is in the matrix clause or the embedded clause.

2.3.45 Summary In this sub-section, we discussed three empirical observations, Adverb Placement, Word Order Inversion and Quantifier Scope provided as support for the thesis that NPAcc occupiesying the position in the matrix clause. Our examination, however, reveals that [all of the arguments do not ==>none of the arguments?] convincingly show that NPAcc behaves as an element of the matrix clause. In the next subsection, we provide two arguments which show the position of NPAcc.

2.4 Valid Argument for NPAcc being in Matrix Clause We have seen in Section 2.3 that many of the arguments provided as a support for NPAcc in the matrix clause are invalid. In this sub-section, we consider two arguments involving idiom expressions and scope, both of which show that NPAcc is an element of the matrix clause.

2.4.12 Idiom Expressions Takano 2003 reports that idiom expressions cannot be laid across NPAcc and the embedded

14 Yukiko Tsuboi Draft of Qual Paper

clause in Japanese whereas they can in English, suggesting that RTO in Japanese and that in English have distinct properties and should not be analyzed in the same way. Idiomatic expressions in English have been known to tolerate raising operation preserving their idiomatic meaning;

(0) Davies and Dubinsky 2004:8: (23a) & (24a) a. The cat seemed to be out of the bag. b. Tina believed the cat to be out of the bag by now.

(0a) is a case of Raising to Subject, and (0b) is a case of Raising to Object, (and?) both of sentences can be interpreted idiomatically. In Japanese, however, idiom interpretation cannot be preserved when idiom expressions extend across NPAcc and the embedded clause in RTO as illustrated in (0).

(0) Takano 2003: 822: (79a) a. John-wa so-ko-made te-ga mawar-anai to omotta. John-TOP that-place-to hand-NOM get:around-not that thought

'John thought that he couldn't take good take care of that.' b. John-wa te-ga so-ko-made mawar-anai to omotta. John-TOP hand-NOM that-place-to get:around-not that thought

c. *John-wa te-o so-ko-made mawar-anai to omotta. John-TOP hand-ACC that-place-to get:around-not that thought

In (0a), te-ga mawaranai 'cannot take care of something' appears as the embedded complement and the idiom interpretation is available there. In (0b), te-ga 'hand-NOM' is dislocated preceding the adjunct, so-ko-made, 'till there'. (0c) is an RTO sentence in which te 'hand' is NPAcc, and the sentence is judged as unacceptable under the intended interpretation. Bruening 2001 reports contradicting judgments in regard to availability of idiomatic interpretation when idiom expressions are laid across NPAcc and the embedded CP;

(0) Bruening 2001: 21: (61) a. Taroo-ga so-no seejika-no kao-o (orokanimo) hiroi to omotta. Taroo-NOM that-GEN politician-GEN face-ACC (stupidly) wide that thought

15 Yukiko Tsuboi Draft of Qual Paper

'Taroo (stupidly) thought that that politician was well-known.' b. Taroo-ga John-no ashi-o chi-ni tuiteinai to omotta. Taroo-NOM John-GEN leg-ACC ground-to reach:not C thought

'Taroo thought that John was restless.' c. Taroo-ga John-no ketsu-o aoi to omotta (koto). Taroo-NOM John-GEN hip-ACC blue-C thought

'Taroo thinks that John is inexperienced.'

Takano 2003 gave different judgments on examples similar to (0), and Yoon 2004 report that three of his informants did not agree with the judgments on the very examples given in (0). Our judgment is on a par with those reported by Takano 2003 and Yoon 2004. We suspect if the judgment in (0) can be shared with native speakers of Japanese. Notice that all of the NPAcc in (0) are preceded by a genitive phrase, so-no seizika-no, John-no and John-no respectively. Adding such modifications seems to make judgments of sentences blurry irrespective of whether a sentence is RTO construction or not. (0) illustrates such point.

(0) a. ?Taroo-ga so-no seejika-no kao-ga hiroi to omotta (koto) Taroo-NOM that-GEN politician-GEN circle:of:aquaintance wide C thought (that)

b. ??Taroo-ga John-no ashi-ga ti-ni tuiteinai to omotta (koto) Taroo-NOM John-GEN leg-NOM ground-to reach:not C thought

c. ??Taroo-ga John-no ketsu-ga aoi-to omotta (koto). Taroo-NOM John-GEN hip-NOM blue-C thought

ここからはじまる話は、(0a)の例が OK になってもおかしくないといっているのだと判断し ましたが、そうなると、(0)のすぐ下で(0a/b/c)すべてを一緒くたにして"Our judgment is on a par with those reported by Takano and Yoon"というところと折り合いが悪い気がします。違和 感をこれ以上うまく説明できませんが。すみません。=>(0a) sounds better compared to (0b) or (0c) because the literal interpretation of the idiom is not readily available because kao 'face' can generally be not hiroi 'wide' or semai 'narrow' but big or small, and hiroi 'wide' forces us to interpret kao 'face' idiomatically. It is should also be noted that the expression, kao-ga hiroi may not be

16 Yukiko Tsuboi Draft of Qual Paper

interpreted as an idiomatic expression but kao itself is. kao in kao-ga hiroi expresses that 'acuqanitance' or 'circle of friends', which can be used in a sentence without hiroi 'wide' preserving idiomatic interpretations as evidenced by (0).

(0) a. ima-no uti-ni kao-o hirogeteoku to, business-ga raku-ni narudaroo. now-GEN inside-in circle:of:friends widen that, business-NOM easy-to become:will

'If you try to widen a circle of friends now, your business would go well later.' b. zimitina doryoku-no okage-de, John-wa kao-ga dan dan hirogattekita. constant efforts-GEN due:to John-top acquanitances-also bit:by:byt became:to:be:wider

'Due to the constant efforts, John's circle of friends has become wider bit by bit. c. ?Tanaka san -wa kao-ga semai node, eigyoo-ni-wa muk-anai. Tanaka MR-TOP circle:of:friends narrow because sales-for-TOP appropriate- NEG

'As for Mr Tanaka, his circle of friends is very narrow, and he is not suitable for sales.'

Instead of hiroi, a transitive verb, hirogeru 'widen', an intransitive verb, hirogaru 'become to be wider" and semai 'narrow' are used in (0a), (0b), and (0c) respectively. Although the acceptability of (0c) is somewhat lower, it is far from being unacceptable. ashi-ga ti-ni tuku used in (0b) cannot readily retain its idiomatic reading when a part of the idiom is modified by genitive phrases despite the fact that the intended meaning is pragmatically natural;

(0) a. *?John-no asi-ga ti-ni tuiteinai (koto) John-GEN leg-NOMTOP ground-at reach:neg (the fact that))

Intended: '(the fact that) John is not stable yet' b. *?John-no ashi-ga ti-ni tuku made-wa nakanaka ansin dekinai naa. John-GEN leg-NOM ground-at reach till-TOP yet relieved cannot SFP

Intended: 'We will not be able to feel relieved till John becomes secure.'

It is also noted that kao can form a number of idiomatic expressions, and such wide variety of usage

17 Yukiko Tsuboi Draft of Qual Paper

can be taken as an indication that the idiomatic use of kao is less restricted than other idiomatic expressions.

(0) a. kao-ga kiku 'well-known' face-NOM work

b. kao-o tubusu 'make one's face lose' face-ACC crush

c. kao-ni doro-o nuru 'make one's reputation lower' face-to dirt-mud spread

d. kao-to naru 'become to be a representative of' face-to become

k. kao-o awaseru 'meet' face-acc put:together f. kao-o dasu 'show up' face-ACC show

g. kao-o naraberu 'members show up' face-ACC line:up

h. kao-ga ureru 'become to be well-known' face-Nom sell

i. kao-o kasu 'see someone for somebody' face-ACC lend

j. kao-o tateru 'keep reputation' face-ACC stand

Thus, in order to conduct more refined experiment, we need to avoid using idiom chunks which contain idiomatic use of an NP and use an idiomatic expression which can have a special meaning only when it is combined as "idiom". (0a)-(0c) are such examples; bati, tujituma and do do not seem to have idiomatic meaning by themselves.

(0) a. bati-ga ataru

18 Yukiko Tsuboi Draft of Qual Paper

'to get punished, to be attacked as revenge' b. tujituma-ga awanai 'not consistent' c. do-ga sugiru 'over do, go too far'

If we place such idiom chucks in RTO constructions, they do not tolerate to extend over NPAcc and the embedded clause as in (0)-(0).

(0) (John's company went out of business because he has been making a profit by deceiving a number of customers. Recalling what he has done, [John thought<=a/b の前にこれは いらないのではないでしょうか?];) a. John-wa bati-ga atatta to omotta. 'John thought that it was a payback.' b. *John-wa bati-o atatta to omotta.

(0) (John heard about the crime from a victim but the story didn't match with what he saw in the crime scene.) a. John-wa tujituma-ga awanai to omotta. 'John thought that it doesn't match.' b. *?John-wa tujituma-o awanai to omotta.

(0) (Bill likes joking but he always goes too far and ends up hurting people by his jokes. Regarding Bill's joke, John thinks:) a. 'John-wa do-ga sugiru-to omotteiru. 'John thinks that Bill goes too far.' b. *John-wa do-o sugiru-to omotteiru.

In every example, the idiomatic expressions laid across NPAcc and the embedded clause lose their idiomatic meanings. Takano 2003 suggested that the situation is similar not to RTO in English but to "proleptic construction". (0) is an example of RTO in English, and (0), "proleptic construction".

(0) Takano 2003: 823: FN 24 (i)

19 Yukiko Tsuboi Draft of Qual Paper

a. John believes the shoe to be on the other foot. b. John believes advantage to have been taken of the workers.

(0) Takano 2003: 822: (81) a. *John believes of the shoe that it's on the other foot. b. *John believes of advantage that it was taken of the workers.

The idiom expressions in (0) can preserve their idiomatic interpretations whereas they do not in "proleptic constructions" in (0). Hoji 1991 claims that NPAcc corresponds to of NP in English, with which Takano's suggestion is in harmony with. If this line of analysis is on the right track, it is plausible to assume that NPAcc in Japanese RTO constructions occupies the same position as the shoe or advantage in (0), which is a position in the matrix clause.

2.4.21 Neg in the embedded clause cannot scope over NPAcc. It is generally agreed that both a QP and a negation are scope bearing element which exhibit scope interactions. As such, a QP and a negation in a sentence of the form in (0) have two interpretations; (i) QP takes scope over negation and (ii) negation takes scope over QP.

(0) [IP [sub QP]...V-neg.]

Such scope ambiguity can be observed in Japanese (Kuno 1980, Hasegawa 1991, 1993, Imani 1993). Observe (0);

(0) [sanzyuunin-izyoo-no gakusei]-ga sensyuu-no kaigi-ni syussekisinakatta. thirty:CL-more:than-GEN student-NOM last:week-GEN conference-at attend:not:past

'More than thirty students did not attend the conference held last week.' OKmore than 30 students > Neg OKNeg > more than 30 students

The QP in the subject position can take scope over Neg, and Neg can take scope over the subject QP. The ambiguity remains even when the sentence in (0) is embedded in another sentence as in

20 Yukiko Tsuboi Draft of Qual Paper

(0).

(0) John-wa [CP [sanzyuunin-izyoo-no gakusei]-ga sensyuu-no kaigi-ni syussekisinakatta]-to John-TOP thirty:CL-more:than-GEN student-NOM last:week-GEN conference-at attend:not:past-C

omotteita. thought

'John thought that more than thirty students did not attend the conference held last week' OKmore than 30 students > Neg OKNeg > more than 30 students

However, when Neg is in the embedded clause, it cannot take a wide scope with respect to a QP in the matrix clause as you can see in (0).

(0) [[5ninizyoo-no sensei]-ga [CP John-ga sensyuu-no kaigi-ni syussekisinakatta to] omotteita.] 5CL:more:than-GEN teacher-NOM John-NOM last:week-GEN conference-at attend:neg:past C thought

'More than five teachers thought that John did not attend the conference held last week.' more than five students teachers > Neg *Neg > more than five studentsteachers

It is suggested in Klima 1964, Takubo 1985, explicitly formulated in Kataoka 2004 that a QP must be in the scope of Neg in order to take a scope under Neg. Following the standard assumption, I also assume that the scope interaction between Neg and a QP is regulated based upon their c- command relation as in (0), cited from Kataoka 2004.

(0) Kataoka 2004:27 (5) The scope of a negation is its c-command domain at LF.

Given (0), the condition in (0) can be [deducted<=deduced?] on wide-scope reading of Neg with respect to QP.

21 Yukiko Tsuboi Draft of Qual Paper

(0) Neg can take a wide scope with respect to a QP only if a QP is in the c-command domain of Neg at LF.

Under the hypothesis that NPAcc is in the matrix clause at LF, the following prediction can be made;

(0) Neg in the embedded clause cannot take a wide scope with respect to NPAcc.

The prediction is born out. As shown in (0), the negation attached to the embedded verb can take scope over the embedded subject in the case of non-RTO sentence in (0a) whereas it cannot take a wide scope with respect to NPAcc in the case of RTO sentence in (0b).

(0) a. John-ga 30ninizyoo-no gakusei-ga meeting-ni konakatta to omotteia (koto) John-NOM thirty:CL:more:than-GEN student-NOM meeting-at come:neg:past C thought (fact)

'John thought that more than thirty students did not come to the meeting.' 30 ninizyoo> NEG, NEG> 30 ninizyoo b. John-ga 30 ninizyoo-no gakusei-o meeting-ni konakatta to omotteia (koto) John-NOM thirty:CL:more:than-GEN student-ACC meeting-at come:neg:past C thought (fact)

'John thought about more than thirty students not to have come to the meeting.' 30 ninizyoo> NEG, *NEG> 30 ninizyoo

Based on this fact, we can conclude that NPAcc must be an element in the matrix clause at LF. There are two possibilities with regard to the position of NPAcc; (i) NPAcc is base-generated in the matrix clause and is not a part of the embedded clause at any point of derivation, or (ii) NPAcc originates in the embedded clause and undergoes "A-movement" which does not exhibit reconstruction effects. The issue will be discussed in Section 2.5.

2.4.3Summary In this sub-section, we have considered two arguments, based on idiom expressions and scope interactions. The former fact suggests that NPAcc occupies the position in the matrix clause

22 Yukiko Tsuboi Draft of Qual Paper

because it behaves the same as a "proleptic" object in English. The latter fact forces us to say conclude that that NPAcc is ("i.e.," or "or" ? ends up being) an element in the matrix clause at some point of derivation. Having said concluded that, we will now consider the issue whether or not NPAcc is base-generated in the matrix clause or raise from the embedded clause in the next sub- section.

2.5 Arguments for Raising of NPAcc We have observed that NPAcc behaves like an element in the matrix clause in the previous sections. We are now in a position to investigate if NPAcc is base-generated in the matrix clause or raise from the embedded clause. The only argument provided in the literature which distinguishes two positions is Proper Binding Condition (PBC) violation effects.

2.5.1 Proper Binding Condition Kuno 1976 first observes that the complement clause of RTO cannot be dislocated to the position preceding NPAcc. (0a) is a canonical order of RTO, and the embedded CP, baka da 'stipid' precedes NPAcc, Tanaka-o in (0b), which results in unacceptability of the sentence.

(0) Kuno 1976: 24 (17b), 35: (66) a. Yamda-wa Tanaka-o [baka da to] omotteita. Yamada-TOP Tanaka-ACC [stupid be C] thought

b. *Yamada-wa bakada to Tanaka-o omotteita. Yamada-TOP [stupid be C] Tanaka-ACC thought

The observation can be generalized as in (0).

(0) The embedded clause in RTO construction cannot be preceded by NPAcc.

Sakai 1998, Bruening 2001, Hiraiwa 2003, Yoon 2004, among others analyzed such examples as violation of Proper Binding Condition (PBC), the definition of which is given in (0).

(0) Proper Binding Condition A trace must be bound. (Fiengo 1974)

23 Yukiko Tsuboi Draft of Qual Paper

Under the raising analysis, NPAcc is considered as being raised from the embedded clause, leaving its trace inside the embedded CP as illustrated in (0a).

(0) The hypothesized structures under the raising analysis

a. NP-TOP..NPi-ACC [CP...ti...V2 that] V1...

b. *NP-TOP...[CP...ti...V2 that]j...NPi-ACC tj V1...

When the CP moves across NPAcc, the sentence will have the structure in (0b) in which the trace of NPAcc is not bound. Recall that the observation solely distinguishes the raising analysis from the [major object analysis<=大文字のところとそうでないところがあります。 ] because other arguments provided by the raising analysis advocates can be compatible with Major Object Analysis as we saw in Section 3.4. Hence, the validation of the generalization in (0) is crucial for the raising analysis. The generalization is, however, challenged by Hoji 1991, further discussed in Takano 2003, Harada 2003. Hoji 1991 points out that the RTO sentence in (0) which areis supposed to exhibit PBC violation is not as bad as a general case of PBC violation in a scrambling construction.

(0) Hoji 1991: 2: (5c) & (5d)

a. ??[ti hoka-no dare yori-mo baka da to] John-ga Billi-o {danteisi / omoikom} ta koto.

b. [ti hoka-no do-no biiru-yori mo umai to]j Santorii-ga [so-no dorai biiru]i-o tj sendensiteiru (koto)

(0a) is a non-RTO complex sentence with no movement involved, and PP sono mura-ni 'in the village' is scrambled to the matrix clause leaving its trace inside the CP in (0b). In (0c), the CP is dislocated to the position preceding sono mura-ni, and the trace of the PP is no longer bound, hence, PBC violation.

(0) Hoji 1991: 2: (6) based on Saito 1987: 309: (20b)

a. John-ga [CP Bill-ga sono mura-ni sundeiru to] {danteisita / omotteiru} (koto)

b. John-ga sono murai-ni [CP Bill-ga ti sundeiru to] {danteisita / omotteiru} (koto)

c. *[ CP Bill-ga ti sundeiru to] John-ga sono murai-ni {danteisita / omotteiru} (koto)

24 Yukiko Tsuboi Draft of Qual Paper

One might say that this is just judgmental variations among speakers. If that is the case, we would expect that those who do not find PBC effects in RTO sentences will not find those in scrambling construction either. Yet, such is not the case as revealed in the next section.

2.5.2 Preliminary Survey on PBC violation As we saw in the previous section, two contradicting judgments have been reported in the literature with regard to sentences of the forms in (0).

(0) * / ??/ok NP1-TOP...[...V2 that]...NP2-ACC.V1...

The generalization put forth is (0) repeated below.

(0) The embedded clause in RTO construction cannot be preceded by NPAcc.

In order to check the validity of (0), we asked 18 native speakers of Japanese (7 linguists and 11 non-linguists) to see how clearly they would find PBC violation as compared to those in scrambling construction. Let us go over the structure of the experiment. We gave each informant the total of 18 sentences which can be divided into two sets. Each set includes four non-RTO sentences and four RTO sentences. Sentences of the form in (0)-(0) are given as examples of non-RTO constructions.

(0) is a non-RTO complex sentence, and NP-DAT/TO or the embedded CP can be dislocated as in (0a) or (0b) respectively.

(0) NP-TOP/NOM...[CP NP-NOM...NP-DAT/TO...V that]...V

(0) a. NPi - DAT/TO ...NP-TOP / NOM... [CP NP-NOM ... ti... V C] ... V

SCRAMBLING

b. [CP NP-  NOM ...NP- DAT/TO ... V C]i ...NP-TOP/ NOM ... ti... V1

SCRAMBLING

When the CP containing the trace of NP-DAT/TO is dislocated crossing NP-DAT/TO as in (0), such

25 Yukiko Tsuboi Draft of Qual Paper

sentence is regarded as unacceptable (Saito 1987) due to PBC violation.

PBC violation configuration in non-RTO

① SCRAMBLING

(0) [CP NP2 -  NOM ... t3...Verb C]i...NP3 - DAT/TO ...NP1- TOP/ NOM... ti... V1

②SCRAMBLING

Sentences of the forms in (0)-(0) are given to the informants as RTO examples. (0) illustrates a canonical order of a RTO sentence with the analysis assumed under the raising analysis.

(0) NP-TOP/NOM...NPi - ACC [CP...ti...V2 C] V1

RAISING

NPAcc can be dislocated as in (0a), and movement of the CP can be followed by the dislocation of NPAcc as in (0b).

②NP-SCRAMBLING

(0) a. NPi-ACC...NP- TOP/NOM...t'i [CP ...ti... V2 C]...V1

①RAISING

③NP-SCRAMBLING

②CP-SCRAMBLING

b. NPi- ACC [CP ... ti... V that]j...TOP/NOM... t'i. tj... V

①RAISING

(0) is a case of PBC violation where NP-scrambling is followed by CP-scrambling which creates an

26 Yukiko Tsuboi Draft of Qual Paper

unbound trace, ti

PBC violation configuration in RTO under the raising analysis

②NP-SCRAMBLING

(0) b [CP .. ti... V2 C]j... NP- ACC ...NP- TOP/NOM.. t'i...tj ... V1

①RAISING

③CP-SCRAMBLING

2.5.2.1 Design of Preliminary Survey and Example Sentences The schematic structures of examples given in the experiment are summarized below.

Schematic Structures of Examples non-RTO constructions

(0) NP-TOP/NOM...[CP NP-NOM...NP-DAT/TO...V that]...V (non-Scrambling)

(0) a. NPi-DAT/TO...NP-TOP/NOM... [CP NP- NOM...ti ...V that]...V (NP-scrambling)

b. [CP NP-NOM...NP-DAT/TO...V that]j ...NP- TOP/NOM... tj...V (CP-scrambling)

(0) *[CP NP-NOM...ti...V that]j ...NPi-DAT/TO...NP-TOP/NOM...tj...V (PBC effects)

RTO constructions

(0) NP-TOP/NOM...NPi-ACC [CP...ti...V that] ...V (non-Scrambling)

(0) a. NPi-ACC...NP-TOP/NOM...[CP ...ti... V that] ... V (NP-scrambling)

b NPi-ACC [CP...ti... V that]j...NP-TOP/NOM...tj... V (CP-scrambling)

(0) */??/ok [CP ...ti... V that]j NPi-ACC...NP-TOP/NOM...V (PBC effects)

If examples of the forms in (0) are unacceptable due to the PBC violation as the raising analysis advocates claims, such examples should be judged as bad as examples of the form in (0) which is a general case of PBC violation. Two sets of examples with different lexical items are given to the informants corresponding to the above structures. (0)-(0) form the first set, and (0)-(0), the second set.

27 Yukiko Tsuboi Draft of Qual Paper

1st set: non-RTO constructions (0) keisatu-wa [John-ga tyuugoku-ni nigeta to] danteisita. police-TOP [John-NOM China-to escaptedCOMP] determined

Intended: 'The police determined that John escaped to China.' (0) a. tyuugoku-ni keisatu-wa [John-ga nigeta to] danteisita. b. [John-ga tyuugoku-ni nigeta to] keisatu-wa danteisita. (0) [John-ga nigeta to] tyuugoku-ni keisatu-wa danteisita.

1st set:RTO-constructions (0) keisatu-wa John-o tyuugoku-ni nigeta-to danteisita. police-TOP John-ACC escapted COMP determined

Intended: 'The police determined that John escaped.' (0) a. John-o keisatu-wa [tyuugoku-ni nigeta to] danteisita. b. John-o [tyuugoku-ni to nigeta to] keisatu-wa danteisita. (0) [tyuugoku-ni nigeta to] John-o keisatu-wa danteisita.

2nd set: non-RTO constructions (0) Yamada sensei -ga [John-ga kaisya-ni syuusyokusita to] omoikondeita(koto) Prof:Yamada -NOM John-NOM Toyota:at got:a:job COMP believed

Indented: 'Prof. Yamada had believed that John got a job at Toyota.' (0) a. Toyota-ni Yamada sensei -ga [John-ga syuusyokusita to] omoikondeita(koto) b. [John-ga Toyota-ni syuusyokusita to] Yamada sensei-ga omoikondeita(koto) (0) [John-ga syuusyokusita to] Toyota-ni Yamada sensei-ga omoikondeita(koto)

2nd set: RTO-constructions (0) Yamada sensei-ga John-o Toyota-ni syuusyokusita to omoikondeita(koto) Prof. Yamada-NOM John -ACC Toyota:at got:a:job COMP believed

Indented: 'Prof. Yamada had believed that John got a job at Toyota.' (0) a. John-o Yamada sensei -ga [Toyota-ni syuusyokusita to] omoikondeita(koto) b. John-o [Toyota-ni syuusyokusita to] Yamada sensei -ga omoikondeita(koto) (0) [Toyota-ni syuusyokusita to] John-o Yamada sensei -ga omoikondeita(koto)

28 Yukiko Tsuboi Draft of Qual Paper

For each sentence, the informants were given a scale illustrated in (0) and asked to judge them by selecting one of the five circles. The five circles were then calculated into -2, -1, 0, +1 or +2, where -2 corresponds to "bad" and +2 corresponds to "good."

(0) Bad< ======>Good o o o o o

2.5.2.2 Results of Survey 18 informants (7 linguists and 11 non-linguists) participated in the experiment, and the result of the experiment is summarized in (0).

(0) Result Chart (a) Non-scrambling (b) NP-scrambling (canonical order) -2 -1 0 +1 +2 -2 -1 0 +1 +2 1. 1st Set: non -RTO (0) 0 0 0 0 18 (0a) 2 3 1 5 7 2. 2nd Set: non-RTO (0) 0 0 2 4 12 (0a) 3 2 2 5 6 3. 1st Set: RTO (0) 0 4 2 2 10 (0a) 0 3 2 3 10 4. 2nd Set: RTO (0) 4 2 1 2 9 (0a) 0 3 2 4 9

(c) CP-scrambling (d) PBC effects -2 -1 0 +1 +2 -2 -1 0 +1 +2 1. 1st Set: non -RTO (0b) 0 0 0 0 18 (0) 14 2 0 2 0 2. 2nd Set: non-RTO (0b) 0 0 0 1 17 (0) 15 1 0 2 0 3. 1st Set: RTO (0b) 3 2 1 6 6 (0) 2 4 3 5 4

29 Yukiko Tsuboi Draft of Qual Paper

4. 2nd Set: RTO (0b) 2 0 3 6 7 (0) 0 7 6 4 1

Column (a) shows the result of non-scrambling, Column (b), NP-scrambling, Column (c), CP- scrambling, and Column (d), PBC effects. Row 1 and Row 2 in each column are the result of non- RTO sentences, Row 3 and Row 4, RTO sentences. Let us first look at Column (a). In the case of non-RTO sentences in Row 1 and Row 2, 18 informants in the first set and 12 in the second set judged examples in (0) and (0) +2 respectively, and nobody judged gave them negative numbers. In contrast, the judgments for the non-scrambling cases of RTO (Row 3 and Row 4 in Column (a)) sentences vary, and four informants in the first set and six informants in the second set even gave (0) and (0) negative numbers. This shows that there are some speakers who do not like RTO sentences even they are in canonical order. Under Column (b), judgmental variations spread even more widely but more than half of informants accept NP scrambling cases of RTO and non-RTO. In the case of CP scrambling in Column (c), all of the non-RTO examples fall into plus range whereas RTO examples show a great deal of judgmental variations. Column (d), which is a case of PBC effects, show a significant difference between non-RTO and RTO (See the shaded part in the chart under Column (d)). Example (0) and (0) are non-RTO examples which are generally considered as exhibiting PBC effects. They are judged as -2 by 14 informants in the first set and 15 informants in the second set. In sum, of all 18 informants, 16 of them gave negative number to such sentences. In the case of RTO examples, however, only 2 informants in the first set and nobody in the second set gave them -2. Even if we add the informants who gave -1, only 6 in the first set and 7 in the second found the sentence unacceptable, the result of which are not very different from those under the non-scrambling case of RTO sentences in Column (a), which are not a case of PBC. Notice that there are two informants who do not find PBC effects even in non-RTO constructions. We conducted a follow-up experiment on one of them. The result of his judgments in the first is summarized in (0).

(0) Result Chart (a) Non-scrambling (b) NP-scrambling (canonical order) 1. 1st Set: non -RTO (0) +2 (0a) +2 2. 2nd Set: non-RTO (0) +2 (0a) +1 3. 1st Set: RTO (0) +2 (0a) +2

30 Yukiko Tsuboi Draft of Qual Paper

4. 2nd Set: RTO (0) +2 (0a) +2

(c) CP-scrambling (d) PBC effects 1. 1st Set: non -RTO (0b) +2 (0) +1 2. 2nd Set: non-RTO (0b) +1 (0) +2 3. 1st Set: RTO (0b) +2 (0) +1 4. 2nd Set: RTO (0b) +2 (0) +1

In the first follow-up experiment, the four more sentences are given, which are variants of the PBC examples in the previous experiment. The only difference is that in the first experiment, both the CP and the NP are dislocated to the sentence-initial position whereas in the follow-up experiment, only CP is fronted leaving a scrambled NP in non-RTO sentences or an NPAcc in RTO sentences in the position proceeded by the matrix subject. (0)-(0) are examples given in the experiment. non-RTO sentences

(0) [John-ga ti nigeta to]j keisatu-wa tyuugokui-ni tj danteisita.

[John-NOM ti escapted COMP] j police-TOP Chinai-to tj determined

(0) [John-ga ti syuusyokusita to]j Yamada sensei-ga Toyotai-ni tj omoikondeita(koto)

[John-NOM ti got:a:job COMP] j Prof:Yamada -NOM Toyota:at tj believed

RTO sentences

(0) [ti tyuugoku-ni nigeta to]j keisatu-wa Johni-o tj danteisita.

[ti China-to escapted COMP]j police-TOP Johni-ACC tj determined

(0) [ti Toyota-ni syuusyokusita to]j Yamada sensei -ga Johni-o tj omoikondeita(koto)

[ti Toyota:at got:a:job COMP j] Prof. Yamada-NOM Johni -ACC tj believed

In non-RTO examples of (0) and (0), ni-marked NPs are scrambled out of the CP to the position immediately preceding CP, and subsequently, the CP is fronted crossing over the matrix subject and the NP-ni. In RTO examples of (0) and (0), the CPs are fronted leaving the NPAcc in the position following the matrix subject. In this follow-up experiment, the informant is able to detect PBC effects in the non-RTO constructions more clearly as evidenced by the number -1 given to (0) and

31 Yukiko Tsuboi Draft of Qual Paper

(0). In contrast, PBC violation in non-RTO constructions in (0) and (0) remains not to be found since he gave each example +1 and 0 respectively. This result of the experiments shows that the PBC effects in RTO construction are not as robust at all as it has been claimed in the literature. Given that the PBC effects have been considered as the strongest piece of evidence adduced for the raising analysis, the result reported above seriously undermines it and in turn provides support for the Major Object analysis, under which the PBC violation of the sort noted above is not predicted in RTO.

2.5.2.3 PBC effects in Korean Similar experiments are conducted on Korean by three Korean students, and the results are remarkably similar to those of the experiments in Japanese. One of the experiments conducted by Y. Cho is designed to compare PBC effects in non-RTO constructions very much in the same way as the experiments in Japanese discussed above. In this section, we will see the results of his experiment of PBC effects in Korean. Among other examples, he gave Korean sentences corresponding to the structures in (0)-(0), where (0) and (0) are non-RTO constructions and (0) and (0), RTO constructions. non-RTO constructions

(0) NP-TOP/NOM...[CP NP-NOM...NP-DAT/TO...V that]...V (non-Scrambling)

(0) *[CP NP-NOM...ti...V that]j ...NP-TOP/NOM.. NPi-DAT/TO...tj...V (PBC effects)

RTO constructions

(0) NP-TOP/NOM...NPi-ACC [CP...ti...V that] ...V (non-Scrambling)

(0) */??/ok [CP ...ti... V that]j...NP- TOP/NOM... NPi- ACC V (PBC effects)

The result of the experiment is summarized in (0). The number of the informants participated in the experiment is 16 for non-RTO constructions and 21 for RTO constructions respectively.

(0) Result Chart 2 (a) Non-scrambling (b) PBC effects (canonical order) -2 -1 0 +1 +2 -2 -1 0 +1 +2

32 Yukiko Tsuboi Draft of Qual Paper

A1:1stSet:nonRTO (0) 1 0 0 1 14 (0a) 14 1 0 0 1 B2:2ndSet:nonRTO (0) 0 0 1 0 15 (0a) 14 0 1 1 0 C3:1st Set: RTO (0) 0 0 0 2 19 (0) 6 4 9 1 1 D4:2nd Set:RTO (0) 0 0 3 7 11 (0) 3 7 3 5 3 E5:3rd SetRTO (0) 0 0 6 6 9 (0) 5 7 5 1 3 F6:4th Set:RTO (0) 3 6 4 3 5 (0) 5 8 3 4 1

As shown in Column (a), 14 informants in the first set and 15 in the second set found non- scrambling examples of non-RTO perfectly acceptable, and almost the same numbers of informants found PBC effects as in Column (b). In contrast, the judgments for the RTO constructions vary a great deal as you can see Row 3-6 in Column (a) and (b). Since about half of the informants marked the examples of the form in (0) with a negative numbers, one may say that there is a contrast between non-RTO constructions and RTO constructions paying attentions only to the numbers in the shaded part of the chart. What is significant, however, is the degradation of acceptability from non-scrambling examples to PBC examples because the RTO examples are not found perfectly acceptable even when they are in canonical order as shown in Column (a). The average score of each example is calculated in (0).

(0) Average Scores of each example (a) Non-scrambling (b) PBC effects (c) Degradation (canonical order)

A:1stSet:nonRTO (0) +1.68 (0a) -1.68 -3.36 B:2ndSet:nonRTO (0) +1.87 (0a) -1.68 -3.55 C:1st Set: RTO (0) +1.90 (0) +0.61 -1.29 D:2nd Set:RTO (0) +1.38 (0) +0.09 -1.29 E:3rd SetRTO (0) +1.14 (0) +0.41 -0.71 F:4th Set:RTO (0) +0.04 (0) +0.57 +0.53

We need to pay attention to how much degradation is observed examples of PBC effects in Column (b) as compared to those of non-scrambling in Column (b) in each non-RTO and RTO constructions. The degradation is calculated Under Column (c). In the case of non-RTO, the

33 Yukiko Tsuboi Draft of Qual Paper

degradation is -3.36 and -3.55 while in RTO, it is only -1.29, -1.29, -0.71, and +0.53. The result shows that the PBC violation in RTO is not as strong as those in non-RTO even in Korean.

2.5.2.4 PBC effects and Inanimate NPs Yoon 2005 observes that there can be more than two NPAccs in an RTO in Korean, one of which can be base-generated and the other of which undergoes movement from the embedded clause. He further claims that if NPAcc is an inanimate NP, it must be generated in the embedded clause and gets raised. Under the analysis, structures of RTO sentences with animate NPAcc is ambiguous between (0a) or (0b) while RTO with inanimate NPAcc should unambiguously corresponds to the structure in (0).

NPAcc is an animate NP.

(0) a. NP-TOP/NOM...NPi-ACC [CP...ti...V that]...V

b. NP-TOP/NOM...NP-ACC [CP...V that]...V NPAcc is an inanimate NP.

(0) NP-TOP/NOM...NPi-ACC [CP...ti...V that] ...V

In the experiments on the PBC discussed in the previous sections, all the examples contain animate NPs as NPAcc. Under the Yoon's analysis, such examples can correspond to the structure in (0) where there is no trace inside the CP, and the lack of the PBC effects are not surprising. Given the hypothesis that RTO examples with an inanimate NP is always derived through movement, the following prediction can be made;

(0) If an NPAcc is an inanimate NP, the embedded clause in RTO construction cannot be preceded by NPAcc.

A. Ueyama conducted an experiment to check the prediction in (0) in spring 2005. 27 informants were given one set of example of the form in (0) and seven sets of examples of the form in (0). The average score of the answers are summarized in (0). non-RTO constructions with an inanimate NP

(0) a. NP-TOP/NOM...[CP NP-NOM...NP-DAT/TO...V that]...V (non-Scrambling)

b. *[CP NP-NOM...ti...V that]j ... NPi-DAT/TO...NP-TOP/NOM..tj..V (PBC effects)

34 Yukiko Tsuboi Draft of Qual Paper

NPAcc is an inanimate NP.

(0) a. NP-TOP/NOM...NPi-ACC [CP...ti...V that] ...V

b. [CP...ti...V that] NPi-ACC...NP-TOP/NOM...V (PBC effects)

(0) Average Scores of each example (a) Non-scrambling (b) PBC effects Degradation (canonical order) non-RTO (0a) +1.96 (0b) -1.52 -3.48 1st Set. RTO (0a) +2.00 (0b) +0.81 -1.19 2nd Set .RTO +2.00 +0.44 -1.56 3rd. Set RTO +2.00 +0.67 -1.33 4th. Set RTO +1.96 +0.89 -1.07 5th Set. RTO +1.96 +0.52 -1.44 6th Set. RTO +2.00 +0.78 -1.22 7th Set. RTO +2.00 +0.22 -1.78

The prediction in (0) is not borne out. The degradation (shaded column in (0)) in RTO cases is not as great as that in non-RTO cases. Most significantly, examples of the form in (0b) were not given negative numbers by any informants, which strongly suggest that PBC in RTO constructions is not observed contra to what the raising analysis advocates have lead us to believe regardless of the type of NPAcc.

2.5.2.5 Summary The preliminary surveys discussed in this Section 3.5 show that (i) PBC effects in RTO constructions do not [exhibit <==I don't think "exhibit" can be used as an unaccusative verb. You need to work on the entire sentence, making "PBC effects" as an object.] as strongly as that in non- RTO constructions (ii) the result does not change regardless of whether NPAcc is animate or inanimate. Such results cannot be accounted for under the raising analysis in which NPAcc is claimed to leave its trace inside the embedded clause. Recall that PBC effects is the only argument which favors the raising analysis over Major Object analysis. The result of the experiments thus devastates seriously undermine the validity of the raising analysis.

35 Yukiko Tsuboi Draft of Qual Paper

[One may wonder what the source of the alleged PBC effects in RTO is. I maintain it is due to breaking "aboutness relation" as suggested in Hoji 1991, followed by Takano 2003 and Harada 2003 (cf. Kuno 1973). Hoji 1991 suggested that NPAcc and CP in RTO constructions is in an" aboutness relation" much as in the case of "aboutness relation" in topic construction as in (0a) where 'that movie' is in an "aboutness relation" with the rest of the sentence.

(0) a. a-no eiga-wa John-ga syutuensita yakusya-o zenin kiratteiru. that-GEN movie-TOP John-NOM played actor-ACC all hate 'As for that move, John hates all the actor who played (in the movie).' b. ??/?*syutuensita yakusya-o a-no eiga-wa John-ga zenin kiratteiru. (cf. Takano 2003) <==The font sizes change here.]

(0b) sounds a little awkward because a-no eiga-wa cannot easily be in "aboutness relation" with the rest of the sentence due to the preposed NP, syutuensita yakusya-o, disconnected from John-ga zenin kiratteiru. Judgments on example s like (0b) vary as "aboutness conditon" is understood as a pragmatic rule much as in the case of the alleged PBC effects in RTO constructions. I thus attribute the alleged PBC effects in RTO to some sort of a pragmatic rule, not to a unbound trace.

2.6 Summary In this section, we first summarized [the proposals of <== Delete?] the raising analysis and the Major Object analysis and evaluate four arguments provided as evidence that NPAcc is in the matrix clause, which are compatible with either analysis. While showing that none of the four arguments are not without a problem, I provide two other arguments which show more convincingly that NPAcc is in fact in the matrix clause at some part of derivation. In the following sub-section (Section 2.5), we discussed the result of the experiments on the PBC violation effects, which is the only argument favoring the Raising analysis over the Major Object analysis and the results of the experiments indicate, quite clearly, successfully concluded that the PBC effects are not as robust as it has been believed in the literature.

3. ECM analysis

We have considered two major analyses, Raising analysis and Major Object analysis, proposed

36 Yukiko Tsuboi Draft of Qual Paper

for RTO constructions in Japanese. In this Section, we will consider the third type of analysis, ECM analysis, which claims that NPAcc can stay inside the embedded clause throughout the derivation. ECM analysis proposed in J.E.Yoon 1989, further elaborated in Bruening 2001, Hiraiwa 2002. J.E. Yoon 1989 does make clear about the position at which NPAcc is generated. She states that NPAcc can be base-generated in or move from other position inside the embedded clause to Spec of CP, which is regarded as 'major subject' position in her analysis. Bruening 2001 propose that NPAcc can either be moved to the CP Spec or be base-generated at Spec of CP. Under the analysis pursued by Hiraiwa 2002, NPAcc undergoes obligatory movement to Spec of CP and it can optionally raise to the matrix clause. (0) is an illustration of the structure proposed in Bruening and Hiraiwa 2002

37 Yukiko Tsuboi Draft of Qual Paper

(0) ...NP1- NOM...[CP NP2-ACC...V2-to]...V1 (that)... TP wo vP T wo

NP1-NOM v' ty VP ty V' ty

CP V1 ty

[NP2-Acc]i C' (=NPAcc) 3 TP -to 6

ti....V2....

There are two crucial arguments provided in support of the ECM analysis; one of which is what Hiraiwa 2002 calls 'indeterminate agreement' and the other of which is occurrence of sentence-final particles in the embedded clause, both of which are evaluated in the next sections.

3.1. Case-markers and Sentence Final particle Kuno 1976 first made an observation that sentence final particles which express speaker's mode or a question particle -ka can be attached to the embedded V of RTO examples. The sentence particles, -zo and -naa, indicate that the information provided by the speaker is new to the hearer.

(0) Kuno 1976: 40: (85a, b) and (86) a. Yamada-wa Tanaka-o bakada zo to omotta. Yamada-TOP Tanaka-ACC stupid SFP that thought

38 Yukiko Tsuboi Draft of Qual Paper

'Yamada thought that Tanaka is a fool.' b. Yamada wa Tanaka-o baka da naa to omotta. Yamada-TOP Tanak-ACC stupid SFP that thought

'Yamada thought that Tanaka is a fool.' c. Yamada-wa Tanaka-o baka ka to omotta. Yamada-TOP Tanak-ACC stupid Q that thought

'Yamada wondered if Tanaka was a fool.' SFP stands for Sentence Final Particle

Hiraiwa extends the observation and reports that sentence-final particles can appear when the embedded clause takes a wa-marked NP, which is generally considered as a topic phrase, but not when it takes a ga-marked NP, as exemplified in (0).

(0) Hiraiwa 2002: 8: (22) Taro-ga Hanako-wa / o / *ga baka-da-naa to omotta. Taro-NOM Hanako-TOP/ACC/*NOM stupid-be SentenceFinalParticle C think

The contrast in (0) is taken as evidence that NPAcc and a wa-marked NP both occupy Spec of CP where a topic phrase is generally understood as residing. Hiraiwa's reasoning, however, cannot be correct since the observation in (0) is not valid in two respects; (i) NPAcc and a topic phrase can co- occur in RTO constructions, and (ii) a sentence-final particle can appear with a nominative marked NP. (0) is the examples which include both NPAcc and a wa-marked NP (both indicated in bold) in the embedded CP.

(0) a. Taroo-ga Hanako-o aitu-wa (gakkoo iti-no) bakada naa to omotta (koto) Taro-NOM Hanak-ACC that:person-TOP (schoo #1-GEN) stupid:be SFP thought (that)

'Taro thought that Mary was the most stupid person in class.' b. Mary-ga John-o aitu-wa (marenimiru) syuusaida zo to omotta (koto) Mar-NOM John-ACC that:person-TOP (extraordinally) smart C thought (that)

'Mary thought that John was extraordinally smart.'

39 Yukiko Tsuboi Draft of Qual Paper

In (0), NPAcc and a topic phrase can both show up in one RTO construction without a problem. If NPAcc and a topic phrase occupy the same position as Hiraiwa reasoned, this fact cannot be accounted for unless he makes an ad-hoc stipulation such that there are two slots for CP spec, which would greatly weaken his reasoning. (0) are the examples where a ga-marked NP occurs with a sentence-final particle, contra to what Hiraiwa reported.

(0) a. Taro-ga Mary de-wa nakute, Hanako-ga kawaii-naa to omotteita (koto) Taro-NOM Mary be-TOP not:and Hanako-NOM pretty-SFP that was thinking (the fact)

'Taro was thinking that not Mary but Hanako is pretty.' b. Mary-wa tunezune John-yori-mo Bill-ga sutekida-naa to omotteita. Mary-NOM always John-from-Focus Bill-NOM attractive-SFP that thought

'Mary always has thought that Bill is more attractive than John.'

The above observation shows that Hiraiwa's claim that NPAcc occupies the same position as the topic phrase does not have an empirical basis. In the next sub-section, we will discuss another argument for the ECM analysis, "indeterminate agreement".

3.2 Argument against Major Object analysis: Indeterminate Agreement Kuroda 1965 first observes that wh-phrases in Japanese can be interpreted as corresponding to "any"-words in English2 when a Q-particle -mo is attached to them as illustrated in (0) and (0).

(0) a. dare-mo 'anybody'

2Hiraiwa 2002 refers to the sequence of a wh-phrase...-mo appearing with negation as NPI. It is misleading to call them because they can occur in affirmative sentences as evidenced by the following example provided by Nisigauchi 1990. (i) Nishigauchi 1990: 126 (17) [NP [S Dare-ga kaita] tegami]-ni mo onazi koto-ga kaiteatta. who-NOM wrote letter -in Q the:same thing-NOM written-be Lit. 'In a letter which anybody wrote, the same thing was written.' 'No matter who wrote a letter, the same thing was written in it.'

40 Yukiko Tsuboi Draft of Qual Paper

who-Q b. nani-mo 'anything' what-Q c. doko-mo 'anywhere' where-Q

(0) Kuroda 1965:93 (80), (81) a. dare-mo hon-o kaw-anakat-ta. who-Q book-ACC buy-NEG-PAST

'No one bought books.' b. John-ga nani-mo kaw-anakatta. John-NOM what-Q buy-NEG-PAST

'John did not buy anything.'

A wh-phrase and a Q-particle can appear separately, and they can still be interpreted as "related" to each other.

(0) Kuroda 1965:93 (82) & 94 (83) a. [kore-made dare-ga kangae-mo si-nakat-ta] aidia this-till who-NOM think-Q do-NEG-PAST idea

'the idea which no one has ever thought of' b. John-wa nani-o ka-oo-to-mo si-na-i. John-TOP what-ACC buy-will-that-Q do-not

'John will not buy anything.'

In (0a), the wh-phrase, dare, is the subject of the relative clause, and a Q-particle is attached to vP inside of it. In (0b), the wh-phrase, nani is in the embedded object position and the Q-particle is attached to the embedded clause. (0) illustrates the same point that a Q-particle does not have to be directly attached to a Q-particle.

(0) Hiraiwa 2002: 3:(7)

41 Yukiko Tsuboi Draft of Qual Paper

a. Taro-ga [DP dare-no hon]-mo yoma-nakat-ta. Taro-NOM Indet.-GEN book-Q read-NEG-PST

'Taro didn't read anyone's book.'

b. Taro-ga [vP dare-wo seme]-mo si-nakat-ta. Taro-NOM Indet.-ACC blame-INF-Q do-NEG-PST

'Taro didn't blame anyone.'

c. Taro-ga [NP [CP dare-ga kai-ta] ronbun]-mo yoma-nakat-ta. Taro-NOM Indet.-NOM write-PST paper-Q read-NEG-PST

'Taro didn't read any paper that anyone wrote.'

It is generally agreed that the occurrence of a wh-phrase is restricted in relation to the position of a Q-particle. Such condition can be stated as in (0).

(0) An indefinite wh-pronoun must be in the c-command domain of Q. (Sakai 1998:498, cf. Hiraiwa 2002: 3:(8)3)

With (0) in mind, let us look at examples in (0) and (0) provided by Sakai 1998 and Hiraiwa 2002 respectively , where NPAcc is a wh-phrase and the Q-morpheme is attached to the embedded CP in RTO construction.

(0) (Sakai 1998: 489: (21))

Masao-ga dare-o [S ti baka-da]-to-mo omot-te-inai. -NOM anyone-ACC a fool-COMP-Q think-STAT-NEG

'For no x, Masao thinks that x is a fool.'

(0) (Hiraiwa 2002: 4: (11)) Taro-ga dare (-no-koto)-wo baka da to-mo omowa-nakat-ta. Taro-NOM Indet.(-GEN-thing)-ACC stupid-CPL C-Q think-NEG-PST

3 Hiraiwa's formulation of the condition is slightly different. He claims that an indefinite wh- phrase cannot be licensed by reconstruction. (i) Hiraiwa 2002: 3:(8) An indeterminate NP must be in the c-command domain of cd(Q) in overt syntax.

42 Yukiko Tsuboi Draft of Qual Paper

'Taro didn't consider anyone to be stupid.'

The acceptable status of (0) and (0) are taken as evidence that the wh-phrases are in the c-command domain of a Q-particle -mo, that is, the embedded clause, at some stage of derivation4. If the observation is correct, it posits a serious problem for Major Object analysis which claims that NPAcc is never be a part of the embedded clause. The validity of (0) however, is challenged by Takano 2002.

(0) Takano 2002: 803: (ii) a. ?Watasi-wa dare-ni koi to mo itteinai5. I-TOP who-DAT come (imperative) that Q said:have:not.

'I haven't said to anyone to come.' b. ?Watasi-wa dare-ni sigoto-o suru to mo yakusokusiteinai. I-TOP who-DAT that job0-ACC do that Q promised:have:not

'I haven't promised anyone to do the job.'

(0) clearly violates the c-command requirement in (0) since the wh-phrase is a indirect object of the matrix verb and a Q-particle -mo is attached to the embedded CP. Yet, the sentences are judged as not totally unacceptable. Yoon 2004 also presents a similar judgment on Korean counterpart of (0).

(0) Yoon 2004: 5: (17a) ?Na-nun nwukwu-eykey [PRO ka-la-ko]-to malha-ci anh-ass-ta. I-TOP who-dat go-imp-comp-Q say-comp not-pst-decl

'I did not tell anyone to go.'

4 Sakai regards the observation as evidence for the raising analysis since indefinite wh-phrase is analyzed as being able to be licensed by reconstruction. In contrast, Hiraiwa 2002 claims that it cannot be licensed by reconstruction, hence (0) is considered as evidence of ECM analysis. We will leave this issue open since availability or unavailability of reconstruction does not affect our discussion here. Recall that our position is that NPAcc is not a part of the embedded clause at any time of derivation.

5 Hiraiwa 2002 reports different judgment on a similar example to this.

43 Yukiko Tsuboi Draft of Qual Paper

If Takano and Yoon are correct, the validity of the condition in (0) will seriously be undermined, and the acceptable status of (0) and (0) will no longer be a problem for Major Object analysis. We conducted a survey to check the validity of (0), which will be discussed in the next sub-section.

3.2.1Design of Preliminary Survey and Example Sentences This survey is conducted to check the validity of the generalization in (0), repeated below, put forth by Sakai (1998) and Hiraiwa (2002).

(0) An indefinite wh-pronoun must be in the c-command domain of Q-particke, -mo. (Sakai 1998:498, cf. Hiraiwa 2002: 3:(8))

(0) can be restated in (0) so as to make a prediction.

(0) If an indefinite wh-pronoun is not c-commanded by a Q-particle, it cannot be interpreted as "any"-words in English.

The schematic structures of examples given in this survey are illustrated below.

(0) 1st Set

a. NP1-NOM...[CP wh- NOM...V2 C]-Q...V1 Neg(that) (c-command)

b. NP1- NOM... wh -ACC...V2 C- Q...V1 Neg(that) (split)

c. *NP1- NOM... wh- ACC...V2 C...V1 Neg(that) (no Q-particle)

(0a) is the structure of examples which satisfies the c-command requirement in (0) because wh- phrase appears in the subject of the embedded CP and a Q-particle is attached to the CP. (0b) is the structure of Sakai or Hiraiwa's examples, in which a wh-phrase occurs in the position of NPAcc and a Q-particle is attached to the embedded CP. (0c) is the structure of control examples which does not have a Q-particle anywhere in a sentence. The 2nd set includes examples which does not appear to satisfy the c-command requirement. See (0) for the schematic structure of the examples.

(0) 2nd Set: No apparent c-command

a. NP1-TOP... wh-DAT...NP2-ACC- Q V Neg (no c-command)

44 Yukiko Tsuboi Draft of Qual Paper

b. NP1-TOP... wh -DAT-Q...NP2- ACC V Neg (adjacent)

c. *NP1-TOP... wh -DAT...NP2- ACC V Neg (no Q-particle)

(0a) is a structure in which a wh-phrase appears in the position of the dative NP and a Q-particle is attached to the accusative NP. Given the widely accepted view that the dative object asymmetrically c-commands the accusative object in Japanese, examples of the form in (0a) is predicted to be unacceptable. In (0b), a Q-particle is attached to the wh-phrase in the dative object position, and in (0c), there is no Q-particle in the structure. In this survey, the informants are given three sets of examples whose schematic structure corresponds to (0) and (0), and asked to judge them by choosing one of the five circles in the same way as the survey on the PBC effects. The five circles were then calculated into -2, -1, 0, +1 or +2, where -2 corresponds to "bad" and +2 corresponds to "good." (0)-(0) corresponds to the structures in (0), and (0)-(0), (0).

1st Set (0) a.John-ga dare-ga bakada to-mo omowanakatta(koto) John-NOM wh-NOM stupid-C-Q think:neg:past (fact)

b. John-ga dare-o bakada to-mo omowanakatta(koto) John-NOM wh-ACC stupid-C-Q think:neg:past (fact)

c. John-ga dare-o bakada to omowanakatta(koto) John-NOM wh-ACC stupid-C think:neg:past (fact)

Intended: 'John did not consider anyone to be a fool.'

(0) (sekkaku yuumee restaurant-ni yattekita-to iu-no-ni) a. Mary-wa (dasareta) do-no ryoori-ga oisii to-mo kanzinakatta. Mary-TOP (served) which-GEN dish-NOM delicious C-Q feel:neg:past

b. Mary-wa (dasareta) do-no ryoori-o oisii to-mo kanzinakatta. Mary-TOP (served) which-GEN dish-ACC delicous C-Q feel:neg:past

c. Mary-wa (dasareta) do-no ryoori-ga oisii to kanzinakatta. Mary-TOP (served) which-GEN dish-ACC delicous C feel:neg:past

Intended: (Although she came to eat at a very famous restaurant,) Mary did not think

45 Yukiko Tsuboi Draft of Qual Paper

any dish (served there) were delicious.

(0) (100-o koeru kazu-no kikaku-ga teiansareta-no-ni) a. Yamada syatyoo-wa dore-ga omosiroi to-mo omowanakatta. Yamada President-TOP which-NOM interesting C-Q think:not:past

b. Yamada syatyoo-wa dore-o omosiroi to-mo omowanakatta. Yamada President-TOP which-ACC interesting C-Q think:not:past

c. Yamada syatyoo-wa dore-o omosiroi to omowanakatta. Yamada President-TOP which-NOM interesting C think:not:past

Intended: '(Although more than 100 proposals are submitted), President Yamada did not think anything was interesting.'

2nd Set (0) (saikin sigoto-ga tatekondeite, John-wa kyoo-de renzoku hatuka-kan, yasumi-mo torazu-ni kinmusiteiru. sikasi, ) a. John-wa kazoku-no dare-ni guti-mo kobosanakatta. John-TOP family-GEN who-DAT complaint-Q say:neg:past

b. John-wa kazoku-no dare-ni-mo guti-o kobosanakatta. John-TOP family-GEN who-DAT-Q complaint-ACC say:neg:past

c. John-wa kazoku-no dare-ni guti-o kobosanakatta. John-TOP family-GEN who-DAT complaint-ACC say:neg:past

Intended: '(He has been working for the past twenty days without taking any day-off. But ) John didn't say any complaint to his family.'

(0) (kyoo-no paatii-no tame-ni OC-kara USC-made enro harubaru yattekitan-no-ni) a. Mary-wa dare-ni koe-mo kake-zuni kaettesimatta. Mary-TOP who-DAT voice-Q say-without go:home:past

b. Mary-wa dare-ni-mo koe-o kakezu-ni kaettesimatta. Mary-TOP who-DAT-Q voice-ACC say-without go:home:past

c. Mary-wa dare-ni koe-o kakezu-ni kaettesimatta.

46 Yukiko Tsuboi Draft of Qual Paper

Mary-TOP who-DAT voice-ACC say-without go:home:past

Intended: (She came all the way from OC to USC to attend today's party but) she went home without talking to anybody.

(0) (John-wa kaisyade-mo hizyoosikina koto-o suru koto-de yuumee-da. kyoomo, ) a. John-wa dare-ni renraku-mo sezu, ikinari kaisya-o yasunda. John-TOP who-DAT contact-Q do:without, suddenly company-ACC missed

b. John-wa dare-ni-mo renraku-o sezu, ikinari kaisya-o yasunda. John-TOP who-DAT-Q contact do:without, suddenly company-ACC missed

c. John-wa dare-ni renraku-o sezu, ikinari kaisya-o yasunda. John-TOP who-DAT contact do:without, suddenly company-ACC missed

Intended: '(He is famous for absurd behavior at work. Today,) John didn't come to work without informing anybody.'

3.2.2Results of Survey 20 native speakers of Japanese (6 linguists and 14 non-linguists) participated in this survey, and the result of the survey is summarized in (0).

(0) The result of Survey (A) (B) (C)

-2 -1 0 +1 +2 -2 -1 0 +1 +2 -2 -1 0 +1 +2

1.C-comamnd (0a) 7 3 2 2 6 (0a) 1 4 3 8 4 (0a) 2 3 1 8 6

2. Split (0b) 0 1 3 3 13 (0b) 0 5 3 2 10 (0b) 1 7 2 0 11

3.NoQ-particle (0c) 10 7 0 2 1 (0c) 11 5 2 1 1 (0c) 14 3 1 1 1

47 Yukiko Tsuboi Draft of Qual Paper

4. No C-c (0a) 4 10 2 2 2 (0a) 3 8 2 4 3 (0a) 0 6 1 8 5

5. C-command (0b) 1 0 0 3 16 (0b) 2 0 0 2 16 (0b) 1 0 0 2 17

6. No Q-particle (0c) 15 4 0 1 0 (0c) 12 6 0 2 0 (0)c 12 6 1 1 0

As shown in Row 2 (Split), Hiraiwa or Sakai's type of examples, where a wh-phrase is in NPAcc position and -mo is attached to the embedded clause, are accepted by a large number of the informants. This result may appear to provide a support for ECM analysis at a glance. If you look at Row 1 and Row 4, however, it will become clear that it is not the case. In Row 1, the c-command requirement is satisfied while it is not satisfied in Row 4. If the c-command requirement in (0) is valid, we would expect Example (0a), (0a) and (0a) to be totally unacceptable but Row 1 and Row 4 show that acceptability of sentences are not affected with or without c-command. Furthermore, another survey conducted by M. Irie in Fall 2004 strongly suggests that the c-command requirement in (0) is invalid. In the survey, 12 native speakers of Japanese are asked to judge 20 examples. Of the 20 examples, three sentences in (0) have the same structure as the examples provided by Takano and Yoon, which do not satisfy c-command condition. (0a) is the very example which Hiraiwa marks as *?.

(0) a. Hiraiwa 2002: 6; (15b) Taroo-ga dare-ni MIT-ni iku koto-mo susumenakatta (koto) Taroo-NOM who-DAT MIT-to go C-Q recommend:not:past

'Taro didn't recommend to anyone to go to MIT.' b. watasi-wa dare-ni koi to-mo itteinai. I-TOP who-dat go C-Q have:said:not

'I haven't said to anyone to come.' c. watasi-wa dare-ni so-no sigoto-o suru to-mo yakusokusiteinai. I-TOP who-dat go that-GEN work-ACC do C-Q promised:have:not

'I haven't promised anyone to do that work.'

In (0a-c), the wh-phrase is in the matrix dative object position which cannot be c-commanded by the Q-particle -mo attached to the embedded CP. (0) is the summary of the judgment reported by the informants.

48 Yukiko Tsuboi Draft of Qual Paper

(0) The result

-2 -1 0 +1 +2

1. (0a) 3 4 1 4 0

2. (0a) 0 1 0 2 9

3. (0a) 0 0 0 3 9

Although (0a), the very example which Hiraiwa gave, is not judged as acceptable by a large number of the informants, the other two examples are given positive numbers by many of them. This result strongly suggests that mere existence of Example (0) and (0) cannot constitute evidence for NPAcc being in the embedded clause, to say the least.

3.2.3Summary The results of the two surveys discussed in this section shows two things; (i) Hiraiwa & Sakai's type of examples are generally judged as acceptable, and (ii) the examples which do not satisfy the c-command requirement are not found totally unacceptable by many of the informants. If we only look at the result consider (i), it seems plausible to conclude that NPAcc can be in the embedded clause of the RTO constructions at some point of the derivation as Hiraiwa and Sakai argued. However, (ii) crucially shows that validity of the c-command requirement in (0) is very much questionable. Hence, (i) cannot be taken as strong supporting evidence for ECM analysis.

3.3 Other Wh...Focus Particle The observation in (0) and (0) is not restricted to wh...mo, it can be reproduced with other Q- particles such as -ka.

(0) Masao-ga dare-o [S ti bakada]-to-ka, (ahoda to-ka) omot-teita. -NOM anyone-ACC fool C-Q stupid C-Q think-STAT-NEG

'Masao had a thought such that someone is fool or someone is stupid. Kuroda 1965 observes that wh-phrases can be followed by -ka which corresponds to or in English, and they are interpreted as "some"-words such as someone or something in English.

(0) Kuroda 1965: 97:(114) and (115)

49 Yukiko Tsuboi Draft of Qual Paper

a. dare-ka-ga hon-o katta. who-Q-NOM book-ACC bought

'Someone bought books.' b. John-ga nani-ka-o katta. John-NOM what-Q-Acc bought

'John bought something.'

Suggested in Kuroda (1965), further discussed in Nisigauchi(1990), wh-phrase...-ka is subject to the c-command condition much as in the case of a wh-phrase... -mo as illustrated in (0).

(0) a. Mary-wa dare-no kuruma-ka-o yuzuriuketa. Mary-TOP who-GEN car-Q-ACC inherited

'Mary took over somebody's car' b. John-wa nan-no siken-ka-o ukeni-itta. John-TOP what-GEN exam-Q-ACC take-went

'John went to take some exam.'

The c-command condition on wh...-mo in (0) can be modified so as to include the wh...-ka case;

(0) An indefinite wh-pronoun must be in the c-command domain of a Q-particle -mo or -ka.

Observe (0) in which the wh-phrase is in the embedded subject position and -ka is attached to the CP.

(0) a. Mary-wa dare-ga bakada to ka, (dare-ga kasikoi to ka) (itumo sonna koto bakari) Mary-TOP who-NOM stupid:be that Q (who-NOM smart C Q) always such thing only

kangaeteita. was:thinking

'Mary was always thinking like someone is stupid or someone is smart.'

50 Yukiko Tsuboi Draft of Qual Paper

b. John-wa nani-ga ii to ka, (nani-ga warui to ka) (issai) iwanakatta. John-TOP what-NOM good C Q (what-NOM bad C Q) (never) say:not:past

'John has never mentioned what is good or what is bad.'

(0) is minimally different from (0) in case-marking of the wh-phrases.

(0) a. Mary-wa dare-o baka da to ka (dare-o kasikoi da to ka) (itumo sonna koto bakari) Mary-TOP who-ACC stupid:be that Q (who-NOM smart C Q) always such thing only

kangaeteita. was:thinking

b. John-wa nani-o iito ka, (nani-o warui-toka) (issai) iwanakatta. John-TOP what-NOM good C Q (what-NOM bad C Q) (never) say:not:past

The wh-phrases in the embedded subject position in (0) and NPAcc position in RTO construction (0) can be interpreted as "some"-words. Very much in the same way as wh...-mo disucssed in the previous section, wh...-ka can also occur without apparent c-command as exemplified in (0).

(0) a. John-wa dare-ni Bill-ga nigeta to-ka tugezuni, zutto damatteita.6 John-TOP who-DAT Bill-NOM run:past C-Q without:telling for:a:long:time kept:silent

'John kept silence, without telling somebody, about Bill that he ran away.' b. John-wa dare-ni Arizona koozyoo-o heisasiro to-ka meireisuru maeni, so-ko-no John-TOP who-DAT Arizona factory-ACC close:imperative C-Q order before that-place-GEN

tatenaosian-o kangae-yootosita. reconstruction-ACC think-tried

6 It is note that -ka and -to, which are generally said to be C, occur simultaneously in one sentence in (0)-(0). Furthermore, the reverse order of -ka and -to is also possible, and it yields interrogative sentences.

51 Yukiko Tsuboi Draft of Qual Paper

'John tried to think about a reconstruction plan before he orders somebody to close the factory in Arizona.'

In (0), the wh-phrase is in the matrix direct object position is not c-commanded by the Q-particle -ka attached to the embedded CP. Yet the wh-phrase can be interpreted as "some"-words. The acceptable status of the sentences under intended reading suggests that the c-command requirement in (0) is not valid in the case of wh...ka as well as wh...mo.

3.4 Special Constituent: Kawazoe 2004 Given the conclusion reached in the previous section, one would naturally wonder how we would account for the widely-accepted hypothesis that wh-phrases must be in the c-command domain of Q. Let us summarize what we have seen till the last section first. In a sentence of the form in (0a), wh-phrases can be understood as "any"-words or "some"- words in English.

(0) a. NP1-TOP... wh-DAT...NP2-ACC- Q V Neg (no c-command)

Based on the observations, we concluded that wh-phrases do not have to be in the c-command domain of Q. In drawing such conclusion, we implicitly assumed that the Q-particle attached to the accusative marked NP could not c-command the dative-marked wh-phrase in (0a) because it is widely assumed that the dative NP asymmetrically c-command the accusative NP. Kawazoe 2004 challenges the thesis and proposes that [NP-cm...NP-cm] can form a constituent. If such is indeed the case, a Q-particle attached to the accusative NP can c-command the wh-phrase marked with the dative, and the acceptable status of sentences of the form in (0a) is, in fact, subject to the c- command condition in (0). We will summarize Kawazoe 2004's proposal in this sub-section. Kawazoe 2004 proposes that NP-cm...NP-cm can be base-generated as a constituent. Observe (0), in which the subject NP and the object NP in each conjunct are argued to form a constituent under Kawazoe's analysis.

(0) Kawazoe 2004: (1) [Taroo-ga ringo (-o mittu)]-to [ziroo-ga mikan-o (futatu)] kau. [Taroo-NOM apple (-ACC 3:CL)]-CONJ [Ziroo-NOM orange-ACC (2:CL)] buy

52 Yukiko Tsuboi Draft of Qual Paper

'Taro buys three apples and Ziroo buys two oranges.'

She proposes that the nominative marked NP and the accusative marked NP in each conjunct forms a constituent headed by an ec as illustrated in (0).

(0) Kawazoe 2004: 82: (37)

VP

pP kau

pP pP to Taroo-ga p' Ziroo-ga p'

ringo-o mikan-o p p

Among other alternative proposals7, Koizumi's proposal is one of the analyses for the conjunctive structures.

A sentence in (0) is analyzed as involving V-raising by Koizumi 2000, in that a verb in both conjuncts undergoes across-the-board-raising, as illustrated in (0).

(0) Koizumi 2000 IP

VP kau

7 There are two other major analyses for a sentence of the form of (0): one is PF deletion under phonetic identity proposed by Fukui & Sakai 2003, and the other is "Surprising Constituent" analysis proposed by Takano 2002. Kawazoe 2004 challenges both analyses and shows the necessity of [NP-cm...NP-cm] being base-generated. For the detail discussion, readers should be referred to Kawazoe 2004.

53 Yukiko Tsuboi Draft of Qual Paper

VP1 VP2 to Taroo-ga V' Ziroo-ga V'

ringo-o mikan-o

ti ti

Kawazoe 2004 pointed out that the sentence in (0), in which the dative NP moves out of the both conjuncts in the across-the-board manner preceding V-raising, could not be accounted for under Koizumi's analysis.

(0) [Tom-ga ringo-o futatu]-to [Bob-ga banana-o sanbon] Mary-ni ageta (koto). [Tom-NOM apple-ACC two] CONJ [Bob-NOM banana-ACC three:CL] Mary-DAT gave (that)

Under the analysis in Koizumi 2000, Mary-ni should be scrambled out of the both VPs connected by -to in the across-the-board manner, and subsequently, VP1 undergoes scrambling crossing over Mary-ni as indicated by an arrow in (0). The resultant structure violates Proper Binding Condition because the traces of Mary-ni is not bound. The offending traces are shaded below:

(0)

IP1

54 Yukiko Tsuboi Draft of Qual Paper

VP1

VP2 VP3

to IP2 Tom-ga V' Bob-ga V'

Mary-nii

ti V' ti V' IP3

ringo-o futatu tv banana-o tv tj

Koizumi 2000 argues that PBC rules out not A-traces but A'-traces which are not bound at LF 8.

Under this view, (0) can be saved if VP1 reconstructs to the lowest IP3. Koizumi, hence, make the following prediction:

(0) When a VP is scrambled crossing over the element that has moved out of the VP, such VP must be interpreted in the position before the scrambling.

Kawazoe observes that Koizumi wrongly predicts Example (0) to be acceptable with the bound variable reading between Toyota-sae 'even Toyota' and soko1 'that place' and Toyota-sae and soko2 respectively

(0) cf.Kawazoe 2004: 21: (14d)

[VP so-ko1-no tokuisaki-ga seikyuusyo-o ittuu] to [VP so-ko2-no torihikiginkoo-ga that-place-GEN customerr-NOM bill-ACC one:CL and that-place-GEN main:bank- NOM

fuwatarituuti-o ittuu] [IP Toyota-ni-sae [IP soofusiteita]

8 Under the view that PBC applies derivationally purued in xxx, xxx, a sentence in (0) is ruled out by PBC.

55 Yukiko Tsuboi Draft of Qual Paper

bounced:check-ACC one:CL Toyota-DAT-even sent

'Its customer sent a bill to even Toyota and its main bank sent a bounced check to even Toyota.'

The derivation of Example (0) is illustrated below.

(0) Kawazoe 2004: 21: (14)

a. [VP so-ko1-no tokuisaki-ga Toyota-ni-sae seikyuusyo-o ittuu soofu-siteita]-to [VP so-ko2- no torihikiginkoo-ga Toyota-ni sae fuwatarituuti-o ittuu soofu-siteita].

across-the-board V-raising

b. [IP [VP so-ko1-no tokuisaki-ga Toyota-ni-sae seikyuusyo-o ittuu tv]-to [VP so-ko2-no

torihikiginkoo-ga Toyota-ni sae fuwatarituuti-o ittuu tv] soofu-siteita].

across-the-board scrambling of Toyota-ni

c. [IP Toyota-ni-saei [IP[VP so-ko1-no tokuisaki-ga ti seikyuusyo-o ittuu tv]-to [VP

so-ko2-no torihikiginkoo-ga ti fuwatarituuti-o ittuu tv] soofu-siteita].

cross-the-board scrambling of VP

d. [IP [VP so-ko1-no tokuisaki-ga ti seikyuusyo-o ittuu tv ] to [VP so-ko2-no torihikiginkoo-ga

ti fuwatarituuti-o ittuu tv soofu-siteita]j [IP Toyota-ni-saei [IP tj soofusiteita]

The resultant structure in (0d) violates PBC condition since the trace of Toyota-ni-sae (the shaded traces) is unbound. However, Koizumi 2000 makes a prediction that the VP is interpreted in the position before scrambling, as represented in (0c), such sentence is not a violation of PBC, hence is

56 Yukiko Tsuboi Draft of Qual Paper

predicted to be acceptable. Yet the prediction is not burned out because (0) is unacceptable with the bound variable reading of Toyota-ni-sae and soko1 and Toyota-ni-sae and soko2 respectively. Koizumi 2000 cannot account for unacceptability of (0)9 3.5 Indeterminate Agreement Revisited Under the analysis of Kawazoe 2004, let us again consider the cases in which the c-command requirement is not satisfied. Observe (0a), repeated below.

(0) a. NP1-TOP... wh-DAT...NP2-ACC- Q V Neg (no c-command)

9 Harada 2002 proposed a different idea for "indeterminate agreement" phenomenon making reference to "intonational phrase", while endorsing an observation by Takano 2002 (example in (0)) that the c-command condition do not appear to be relevant for "indeterminate agreement" phenomenon. (i) is the licensing condition on "indeterminate agreement" proposed by Harada 2002: (i) NPIs in Japanese are sensitive to intonation; Daremo 'what-all' should be read with a flat accent pattern in order to be interpreted as an NPI.

(ii) is more general statement of the licensing condition in (i). (ii) A wh-phrase must be in the same "intonational phrase" with a Q in order to be interpreted as "any NP" in English.

Under Harada's analysis, the contrast in (iii) provided originally by Hiraiwa 2002 in support of ECM analysis, is attributed to distinct "intonational phrases" which dare and mo belong to. (iii) illustrates intonational phrases a wh-phrase and a QP belong to. (iii) Harada 2002: 17:(29) without accent markers a. Taroo-ga orokanimo dare-o baka-da to-mo omow-anakatta. Taroo-NOM stupidly who-ACC fool-be C-Q think-neg:past 'Stupidly, Taro didn't consider anyone to be a fool.' b. *Taroo-ga dare-o orokanimo baka-da to-mo omow-anakatta.

In the case of (iiia), where the matrix adverb, orokanimo 'stupidly', precedes a wh-phrase, dare-o 'who-acc', NPAcc and the embedded CP falls into the same intonational phrase while in(iv), where the matrix adverb appears in between NPAcc and the embedded CP, NPAcc and the adverb is in one "intonational phrase" and the embedded predicate and a Q-particle attached to it is in another one. (iv) Harada 2002: 17:(29) a. Taˆroo-ga oˆrokanimodare-o baˆka-da to-mo omow-anakatta. Taroo-NOM stupidly who-ACC fool-be C-Q think-neg:past 'Stupidly, Taro didn't consider anyone to be a fool.' b. *Taˆroo-ga dare-o orˆokanimo baˆka-da to-mo omow-anakatta. Although there are some judgmental variations, the relevant reading is significantly easier to obtain when the licensing condition in (ii) is met. It may be the case that when a special constituent is formed, it has an impact on intonational phrase. I leave this issue for further investigation.

57 Yukiko Tsuboi Draft of Qual Paper

If [wh-DAT...NP2-ACC] in (0a) can be base-generated as a constituent, (0a) could have a structure illustrated in (0), in that wh-phrase is c-commanded by a Q-particle.

(0) NP1-TOP... [wh-DAT...NP2-ACC]-Q V Neg

It is no longer surprising that the wh-phrase in (0a) can be interpreted as "some"-words or "any"- words as it is, in fact, in the c-command domain of a Q as illustrated in (0). Let us go back to RTO construction. If Kawazoe's analysis can be extended to the case of NP- cm..CP-cm, we expect that it is possible for NPAcc in RTO and the embedded CP to form a constituent as illustrated in (0).

(0) Mary-ga [John-o bakada-to] omotteiru (koto) Mary-NOM John-ACC stupid:be-that think (that)

Such being the case, Ex (0), (0), (0) and (0) which appear to violate the c-command requirement do in fact satisfies it. Observe (0) again. Under Kawazoe 2004's proposal, dare-o and bakada-to can be base-generated as a constituent, small P (pP). Assuming that -mo is attached to the small pP, (0) will have the structure in (0).

(0) (Hiraiwa 2002: 4: (11)) Taro-ga dare-o baka da to-mo omowa-nakat-ta. Taro-NOM Indet.(-GEN-thing)-ACC stupid-CPL C-Q think-NEG-PST

'Taro didn't consider anyone to be stupid.'

(0) Taro-ga [dare-o baka da to]-mo omowa-nakat-ta.

TP wo NegP T 3 -i vP -na wo

58 Yukiko Tsuboi Draft of Qual Paper

Taro-ga v' ty VP v ty pP omottei ty pP -mo 3 Dare-o p' 3 CP p ty TP -to 6 bakada

In this structure, the wh-phrase is c-commanded by the Q-particle attached to pP satisfying the c- command requirement. Thus, it is not surprising at all that the wh-phrase can be interpreted as "any"-words. Recall that we have observed other cases such as (0a) or (0) repeated below, where wh-phrases can be interpreted as "any" or "some"-words despite the fact that a Q-particle does not appear to c-command them. They can also be accounted for in the same line of analysis. (0a) (0a) are supposed to have the structures in (0) and (0) respectively under the analysis.

(0a) John-wa kazoku-no dare-ni guti-mo kobosanakatta. John-TOP family-GEN who-DAT complaint-Q say:neg:past

(0) John-wa [pP kazoku-no dare-ni guti]-mo kobosanakatta.

(0) a. Hiraiwa 2002: 6; (15b) Taroo-ga dare-ni MIT-ni iku koto-mo susumenakatta (koto) Taroo-NOM who-DAT MIT-to go C-Q recommend:not:past

'Taro didn't recommend to anyone to go to MIT.'

59 Yukiko Tsuboi Draft of Qual Paper

(0) Taroo-ga [pP dare-ni MIT-ni iku koto]-mo susumenakatta (koto)

In each case, the Q-particle -mo is attached to pP which contains the wh-phrases in it. Such configuration satisfies the c-command requirement on wh-phrases.

3.6 wh-interrogative We have observed the cases where wh-phrases appearing in the position of NPAcc can be interpreted as "any"-words or "some"-words, being related to -mo or -ka immediately following the embedded CP. The same kind of observation can be made with wh...-ka which is interpreted as indirect wh-questions. In this case, the complementizer, -to, can or cannot co-occur with -ka depending on the choice of the matrix verb.

(0) a. John-wa dare-ga hannin-ka (*to) siranai. John-TOP who-NOM criminal-Q (*C) know:not

b. John-wa dare-o hannin-ka (*to) siranai. John-TOP who-ACC criminal-Q (*C) know:not

'John doesn't know who is criminal.'

(0) a. keisatu-wa (tuini) doko-ga azito-ka (*to) tukitometa. police-TOP (finally) where-NOM hideout-Q (*C) find out

'The polce (finally) find out where the hideout is.' b. keisatu-wa (tuini) doko-o azito-ka (*to) tukitometa. police-TOP (finally) where-ACC hideout-Q (*C) find out

(0) a. John-wa dare-ga kyoohan (daroo)-ka to kangaeta. John-TOP who-NOM accomplice (probably)-Q C thought

'John wondered who would be an accomplice.' b. ?John-wa dare-o kyoohandaroo-ka to kangaeta. John-TOP who-ACC accomplice (probably)-Q C thought

(0) a. John-to Tom-wa dare-ga ziki syatyoo-ni fusawasii-ka to gironsita.

60 Yukiko Tsuboi Draft of Qual Paper

John-and Tom-TOP who-NOM next president-for suitable-Q C discussed

'John and Tom discussed who would be suitable for the next President.' b. ?John-to Tom-wa dare-o ziki syatyoo-ni fusawasii-ka to gironsita. John-and Tom-TOP who-ACC next president-for suitable-Q C discussed

The observation above is consistent with what we have seen earlier with respect to wh..-mo or wh...- ka. If Grammar allows NPAcc and CP to form a constituent, a wh-phrase in NPAcc can be c- commanded by -ka appearing immediately after the embedded clause, satisfying the c-command requirement.

3.7 Summary and Consequence We have reevaluated the two piece of evidence in this section provided as a support of ECM analyses; the availability of co-occurrence of sentence-final particles with NPAcc and a topic NP and "indeterminate agreement" which are subject to the c-command requirement. We [show <==You should perhaps follow "have evaluated" in the first sentence in this para and use the present perfect here and elsewhere in the para, to be consistent.] that the former observation cannot be correct based on the fact that not only NPAcc or a topic phrase but also NPAcc together with a topic phrase can co-occur with a sentence final particle, which clearly shows that NPAcc cannot be located in the same position as a topic phrase. Another piece of evidence provided for ECM analyais is "indeterminate agreement,", our investigation of which shows that there are cases where wh-phrases do not appear to obey the c-command requirement which is widely accepted thesis <== A requirement is not a thesis…] cross-linguistically. We then suggest that the c-command requirement is valid but the assumption that NPAcc and the embedded CP do not form a constituent is not correct, by extending Kawazoe 2004's analysis on NP-cm...NP-cm to NPAcc and the embedded CP in RTO. Once Grammar allows a sequence of NP-cm...CP to become a constituent, Hiraiwa or Sakai types of examples in (0) and (0) would no longer posit a problem for Major Object analysis since the wh-phrase in the matrix clause can be c-commanded by a Q-particle appearing immediately following the embedded CP. This line of thinking can receive a support from cleft constructions with two NPs appearing in "focus position, where only one constituent would generally appear, example of which is given in (0).

(0) a. John-ga ageta-no-wa Mary-ni hana-o da.

61 Yukiko Tsuboi Draft of Qual Paper

John-NOM gave-GEN-TOP Mary-DAT flower-ACC be.

Lit. 'It is flowers to Mary that John gave.' b. Mary-ga katta-no-wa Hanako-ni ringo-o da. Mary-NOM bought-GEN-TOP Hanako-to apple-ACC be

Lit 'It is an apple for Hanako that Mary bought.'

NPAcc and the embedded CP in RTO constructions can also appear in the "focus" position of the cleft constructions;

(0) a. John-ga omotta-no-wa Mary-o tensaida-to da. John-NOM thought-GEN-TOP Mary-ACC genius:be-C be

Lit.'It is Mary to be a genius that John thought.' b. ?keisatu-ga zutto sinziteita-no-wa Bill-o siro-to datta. police-NOM long:time believed-GEN-TOP Bill-ACC innocent-C be

Lit. 'It was Bill to be innocent that the police had been believing for a long time.'

It may sound natural when more than two combinations of NPAcc and the embedded CP appear in the "focus" position:

(0) a. keisatu-ga (kyoo madeni) danteisita-no-wa John-o haninda-to, kyooki-o naihuda-to police-NOM (today by) determined-GEN TOP John-ACC criminal-C (weapon- ACC nifebe-C

da. be. Lit. 'It is John to be a criminal (and the weapon to be a knife) that the police has determined.' b. keisatu-ga happyoosita-no-wa dare-o haninda-to-ka doko-o azitoda-to-ka da. police-NOM announced-GEN-TOP who-ACC criminal:be-C-Q where-ACC hideout:be -C-Q be

Lit. 'It is who to be a criminal and where to be a hideout that the police has determined.'

62 Yukiko Tsuboi Draft of Qual Paper

The acceptability of the cleft sentences above can be naturally accounted for under the hypothesis that NPAcc and CP forms a constituent as we have suggested in Sec. 4.4.

##############Harada 2002 intonation 話を追加する。#######################

4. Conclusion

In this paper, we defend Major Object Analysis by presenting empirical evidence (i) that NPAcc behaves as an element in the matrix clause based on the scope interaction NPAcc cand negation in the embedded clause, and (ii) that idiom expressions cannot extend across NPAcc and the embedded CP without losing its idiomatic meaning, which pattern with proleptic constructions in English where a proleptic object is base-generated in the matrix clause. We have also reevaluated by conducting preliminary surveys the two crucial arguments, based on the PBC effects and "indeterminate agreement," provided in support of (the?) Raising Analysis and (the?) ECM analysis respectively. The result of the surveys on PBC violation effects, which have been widely accepted to exhibit in RTO construction, shows that it is not as robust as it has been reported. Especially, [examples of (the alleged) PBC violations in RTO constructions is far more acceptable than "genuine" PBC effects in non-RTO constructions <== I don't think you can say "examples are more acceptable than effects."] , the result which seriously undermine the hypothesis that NPAcc undergoes "raising" in RTO constructions in Japanese. [ECM analysis provides two pieces of arguments <==Redo.] , [sentence-final particles and "indeterminate agreement" <== You seem to use these as if they referred to 'arguments' but it is not clear if you can do that (unless you declare that you do).]. [The former of which is based on a false generalization as discussed in Section 4.1, and the other of which latter ? involves special constituency mechanism as proposed in Kawazoe 2004. Here too the English exposition is somewhat loose. "The former" is an argument, I guess. So, "the argument" is based on a generalization. Fine. But if "the latter" is also an argument, then you are saying that the argument involves special constituency mechanism. Can you really say that?] Following Takano 2002, we point out that an NP which is clearly outside of the c-command domain of a Q-particle attached to the embedded CP can behave as if it is c-commanded by xxx. We suggest the possibility that an element in the matrix clause can be in the c-command domain of a Q-particle immediately preceded by the embedded CP by introducing adopting Kawazoe's analysis that allows NP-cm...NP-cm to form a constituent.

63 Yukiko Tsuboi Draft of Qual Paper

One would naturally wonder how Kawazoe's proposed mechanism is constrained and also wonder what the consequenes are of adopting it, beyond the RTO-related phenomenon. To the extent that you are not in a position to provide a (fully) satisfactory answer to such questions, it is very important that you make sure that the reader will get the following points.

(1) Empirical observations lead us to conclude that either (i) the c-command requirement in question (i.e., for an 'indeterminate pronoun' to be c-commanded by a 'scope marker') must be abandoned or (ii) some form of reanalysis must be assumed to be possible in cases where the c- command requirement appears not to be satisfied.

(2) The empirical observations in (1) are independent of RTO constructions.

(3) You choose to pursue (1-ii) rather than (2-i) because xxx, and the Kawazoe proposal provides us with a concrete way to implement the idea in (1-ii).

The point of the above is that it is not crucial that we adopt the Kawazoe proposal. But if we did not, we would have to assume some other mechanism instead, as long as we do not want to puruse (1-i).

Provided that the ECM analysis do not have any other supporting evidence, the fact seriously weakens their claim that NPAcc can stay inside the embedded clause. Based on the discussion in this paper, we are forced to conclude that NPAcc is never a part of the embedded clause at any time point of derivation. Lastly, it is note that NPAcc can appear when an adverb instead of CP complement occurs as in (0).

(0) a. John-ga Bill-no seikaku-o akaruku omotta (koto) John-nom Bill-Gen personality-ACC cheerful thought (that)

'John thought about Bill's personality to be cheerful.' b. John-ga Mary-no taido-o tumetaku kanzita (koto) John-NOM Mary-GEN attitude-ACC coldly felt (that)

'John felt that Mary's attitude is cold.'

64 Yukiko Tsuboi Draft of Qual Paper

akaruku or tumetaku are generally understood as adverbs which can be supported by examples such as (0) where such adverbs occur with the verbs, aisatusuru 'to greet' and attaru 'to treat'. Given the fact that the verbs in (0) do not take clauses as their complement, it would be implausible to assume akaruku or tumetaku are clauses equivalent to to-clauses in RTO constructions.

(0) a. John-wa Mary-ni akaruku aisatusita. John-TOP Mary-DAT cheerfully greeted

'John greeted Mary cheerfully.' b. John-wa Mary-ni tumetaku atatta. John-TOP Mary-DAT coldly treated

'John treated Mary coldly.'

Given the fact that examples in (0) are acceptable, Grammar should allow omou 'to think' or kanziru 'feel', both of which can be used in the matrix clause in RTO constructions, to co-occur with an accusative marked NP independently of RTO constructions, which suggests that NP-acc can optionally show up with the verbs which do not take an accusative marked NP solely. Incidentally, 'preposing' of the adverbs in (0) results in degradation of acceptability as in (0), much as in the case of the alleged PBC effects discussed in Section 3.5.

(0) a. ??/?* akaruku John-ga Bill-no seikaku-o omotta (koto) b. ??/?* tumetaku John-ga Mary-no taido-o kanzita (koto)

This is not surprising if the alleged PBC effects in RTO constructions are due related to "aboutness condition" as suggested in Section 3.5.2.5 because the adverbs in (0) express characteristics about Bill-no seikaku 'xxx' or Mary-no taido 'xxx' respectively, where the adverbs and the accusative phrase can be understood as being in an "aboutness relation very much in the same way as NPAcc and the embedded CP in RTO constructions. "Indeterminate agreement" can also be observed if Bill and Mary in (0) are replaced by dare 'who' and a Q-particle mo is added to the adverbs in (0).

(0) a. ?/?? John-ga (ko-no tiimu-no) dare-no seikaku-o (tokubetu) akaruku-mo

65 Yukiko Tsuboi Draft of Qual Paper

John-nom (this-GEN team-GEN) who-GEN personality (specially) cheerful-Q

omowanakatta(koto) think:not:past (that) 'John thought about personalities of nobody on this team was specially cheerful.' b. ?/?? John-ga dare-no taido-o tumetaku-mo kanzinakatta (koto) John-NOM who-GEN attitued-ACC coldly-Q feel:not:past (that)

'John felt that nobody's attitude was cold.'

This fact is also compatible with our suggestion in Section 4.4 that Japanese allows NP-cm...CP-cm to form a constituent if the idea can be extended to the case of NP-cm and an adverb.

5. References (to be completed)

Harada, Naomi. (2003) "Raising to Object is NOT an edge phenomenon," Paper presented at the January 2003 Annual Meeting of the Linguistics Society of America, unpublished manuscript, ATR International. Hiraiwa, Ken. (2002) "Raising and indeterminate agreement," Second Draft (June 2002) (The document, states "A revised version to appear in the Proceedings of WCCFL 21," but the paper is not included in the proceedings.). Hoji, Hajime. (1991). "Raising-to-object, ECM and the major object in Japanese," A talk given at Japanese Syntax Workshop at University of Rochester. Hong, Kisun. (1990) "Subject-to-object raising in Korean," in Katarzina Dziwirek, Patrick Farrell, and Elias Mejias-Bikandi (eds.), Grammatical relations: A cross-theoretical perspective, 215- 225, Stanford: CSLI. Kuno, Susumu. (1976) "Subject Raising," in Shibatani, ed., Syntax and Semantics: Japanese Generative Grammar, Academic Press. Marantz, Alec. (1983) "Raising and Category Types in Japanese," in Y. Otsu, et al. eds., Studies in Generative Grammar and Language Acquisition, ICU, Tokyo. Saito, Mamoru. (1983) "Comments on the Papers on Generative Syntax," in Y. Otsu, et al. eds., Studies in Generative Grammar and Language Acquisition, ICU, Tokyo. Sakai, Hiromu. (1998) "Raising asymmetry and improper movement," in N. Akatsuka, H. Hoji, S. Iwasaki, S.-O. Sohn, and S. Strauss (eds.), Japanese/Korean Linguistics 7, 481-497, CSLI,

66 Yukiko Tsuboi Draft of Qual Paper

Stanford. Yoon, James. (to appear) "Raising Specifiers. A Macroparametric Account of Subject-to-Object Raising in Some Altaic Languages," The proceedings of the Formal Altaic Linguistics. Yoon, James. (2004) "Raising and Prominence," a 48-page handout for the talk given at Language Education Institute, Seoul National University, 7/26/04. Yoon, J. E. (1989) "ECM and Multiple Subject Constructions in Korean," in S. Kuno, et. al. eds., Harvard Studies in Korean Linguistics 3.

67

Recommended publications